Bump version v0.6.2

Add option to mute audio playback for file input

.github/workflows/ci.yml

@@ -1,4 +1,4 @@
-name: CI
+name: Test & Build CI/CD
 
 on:
   push:
@@ -7,46 +7,192 @@ on:
     tags:
       - v*
   pull_request:
-    branches:
-      - main
+    branches: [ main ]
+    types: [opened, synchronize, reopened]
 
 jobs:
-  build-and-push-package:
-    runs-on: ubuntu-latest
+  run-tests:
+    runs-on: ubuntu-22.04
+    strategy:
+      matrix:
+        python-version: [3.8, 3.9, '3.10', 3.11, 3.12]
+    steps:
+      - uses: actions/checkout@v2
+      
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v2
+        with:
+          python-version: ${{ matrix.python-version }}
+
+      - name: Cache Python dependencies
+        uses: actions/cache@v2
+        with:
+          path: |
+            ~/.cache/pip
+            !~/.cache/pip/log
+          key: ${{ runner.os }}-pip-${{ matrix.python-version }}-${{ hashFiles('requirements/server.txt', 'requirements/client.txt') }}
+          restore-keys: |
+            ${{ runner.os }}-pip-${{ matrix.python-version }}-
+      
+      - name: Install system dependencies
+        run: sudo apt-get update && sudo apt-get install -y portaudio19-dev
+      
+      - name: Install Python dependencies
+        run: |
+          python -m pip install --upgrade pip
+          pip install -r requirements/server.txt --extra-index-url https://download.pytorch.org/whl/cpu
+          pip install -r requirements/client.txt
+      
+      - name: Run tests
+        run: |
+          echo "Running tests with Python ${{ matrix.python-version }}"
+          python -m unittest discover -s tests
+  
+  check-code-format:
+    runs-on: ubuntu-22.04
+    strategy:
+      matrix:
+        python-version: [3.8, 3.9, '3.10', 3.11, 3.12]
 
     steps:
-    - name: Check Out Repository
-      uses: actions/checkout@v2
+      - uses: actions/checkout@v2
 
-    - name: Set up Python
-      uses: actions/setup-python@v2
-      with:
-        python-version: 3.8
-    
-    - name: Set up FFmpeg
-      uses: FedericoCarboni/setup-ffmpeg@v2
-    
-    - name: Install Additional requirements
-      run: |
-        sudo apt-get -y install portaudio19-dev wget
-      shell: bash
+      - name: Set up Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v2
+        with:
+          python-version: ${{ matrix.python-version }}
 
-    - name: Install Client Requirements
-      run: pip install -r requirements/client.txt
+      - name: Install dependencies
+        run: |
+          python -m pip install --upgrade pip
+          python -m pip install flake8
 
-    - name: Install Server Requirements
-      run: pip install -r requirements/server.txt
+      - name: Lint with flake8
+        run: |
+          # stop the build if there are Python syntax errors or undefined names
+          flake8 . --count --select=E9,F63,F7,F82 --show-source --statistics
+          # exit-zero treats all errors as warnings. The GitHub editor is 127 chars wide
+          flake8 . --count --exit-zero --max-complexity=10 --max-line-length=127 --statistics
 
-    - name: Install Wheel for build 
-      run: pip install wheel twine
-    
-    - name: Build wheel
-      run: |
-        python setup.py sdist bdist_wheel
-    
-    - name: Push package on Test PyPI
-      if: github.event_name == 'push' && startsWith(github.ref, 'refs/tags')
-      uses: pypa/gh-action-pypi-publish@release/v1
-      with:
-        user: __token__
-        password: ${{ secrets.PYPI_API_TOKEN }}
+  build-and-push-docker-cpu:
+    needs: [run-tests, check-code-format]
+    runs-on: ubuntu-22.04
+    if: github.event_name == 'push' && (github.ref == 'refs/heads/main' || startsWith(github.ref, 'refs/tags/'))
+    steps:
+      - uses: actions/checkout@v2
+      
+      - name: Log in to GitHub Container Registry
+        uses: docker/login-action@v1
+        with:
+          registry: ghcr.io
+          username: ${{ github.repository_owner }}
+          password: ${{ secrets.GHCR_TOKEN }}
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v1
+
+      - name: Build and push Docker image
+        uses: docker/build-push-action@v2
+        with:
+          context: .
+          file: docker/Dockerfile.cpu
+          push: true
+          tags: ghcr.io/collabora/whisperlive-cpu:latest
+  
+  build-and-push-docker-tensorrt:
+    needs: [run-tests, check-code-format]
+    timeout-minutes: 60
+    runs-on: ubuntu-22.04
+    if: github.event_name == 'push' && (github.ref == 'refs/heads/main' || startsWith(github.ref, 'refs/tags/'))
+    steps:
+      - uses: actions/checkout@v2
+
+      - name: Log in to GitHub Container Registry
+        uses: docker/login-action@v1
+        with:
+          registry: ghcr.io
+          username: ${{ github.repository_owner }}
+          password: ${{ secrets.GHCR_TOKEN }}
+
+      - name: Docker Prune
+        run: docker system prune -af
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v1
+
+      - name: Build and push Docker GPU image
+        uses: docker/build-push-action@v2
+        with:
+          context: .
+          file: docker/Dockerfile.tensorrt
+          push: true
+          tags: ghcr.io/collabora/whisperlive-tensorrt:latest
+
+  build-and-push-docker-gpu:
+    needs: [run-tests, check-code-format, build-and-push-docker-cpu]
+    timeout-minutes: 20
+    runs-on: ubuntu-22.04
+    if: github.event_name == 'push' && (github.ref == 'refs/heads/main' || startsWith(github.ref, 'refs/tags/'))
+    steps:
+      - uses: actions/checkout@v2
+
+      - name: Log in to GitHub Container Registry
+        uses: docker/login-action@v1
+        with:
+          registry: ghcr.io
+          username: ${{ github.repository_owner }}
+          password: ${{ secrets.GHCR_TOKEN }}
+
+      - name: Docker Prune
+        run: docker system prune -af
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@v1
+
+      - name: Build and push Docker GPU image
+        uses: docker/build-push-action@v2
+        with:
+          context: .
+          file: docker/Dockerfile.gpu
+          push: true
+          tags: ghcr.io/collabora/whisperlive-gpu:latest
+
+  publish-to-pypi:
+    needs: [run-tests, check-code-format]
+    runs-on: ubuntu-22.04
+    if: github.event_name == 'push' && startsWith(github.ref, 'refs/tags')
+    steps:
+      - uses: actions/checkout@v2
+
+      - name: Set up Python 3.8
+        uses: actions/setup-python@v2
+        with:
+          python-version: 3.8
+
+      - name: Cache Python dependencies
+        uses: actions/cache@v2
+        with:
+          path: |
+            ~/.cache/pip
+            !~/.cache/pip/log
+          key: ubuntu-latest-pip-3.8-${{ hashFiles('requirements/server.txt', 'requirements/client.txt') }}
+          restore-keys: |
+            ubuntu-latest-pip-3.8-
+
+      - name: Install system dependencies
+        run: sudo apt-get update && sudo apt-get install -y portaudio19-dev
+
+      - name: Install Python dependencies
+        run: |
+          pip install -r requirements/server.txt
+          pip install -r requirements/client.txt
+          pip install wheel
+
+      - name: Build package
+        run: python setup.py sdist bdist_wheel
+
+      - name: Publish package to PyPI
+        uses: pypa/gh-action-pypi-publish@release/v1
+        with:
+          user: __token__
+          password: ${{ secrets.PYPI_API_TOKEN }}

Audio-Transcription-Chrome/background.js

@@ -157,7 +157,8 @@ async function startCapture(options) {
           multilingual: options.useMultilingual,
           language: options.language,
           task: options.task,
-          modelSize: options.modelSize
+          modelSize: options.modelSize,
+          useVad: options.useVad,
         },
       });
     } else {

Audio-Transcription-Chrome/options.js

@@ -99,7 +99,8 @@ async function startRecord(option) {
           uid: uuid,
           language: option.language,
           task: option.task,
-          model: option.modelSize
+          model: option.modelSize,
+          use_vad: option.useVad
         })
       );
     };

Audio-Transcription-Chrome/popup.html

@@ -15,6 +15,10 @@
     <input type="checkbox" id="useServerCheckbox">
     <label for="useServerCheckbox">Use Collabora Whisper-Live Server</label>
   </div>
+  <div class="checkbox-container">
+    <input type="checkbox" id="useVadCheckbox">
+    <label for="useVadCheckbox">Use Voice Activity Detection</label>
+  </div>
   <div class="dropdown-container">
     <label for="languageDropdown">Select Language:</label>
     <select id="languageDropdown">

Audio-Transcription-Chrome/popup.js

@@ -4,6 +4,7 @@ document.addEventListener("DOMContentLoaded", function () {
   const stopButton = document.getElementById("stopCapture");
 
   const useServerCheckbox = document.getElementById("useServerCheckbox");
+  const useVadCheckbox = document.getElementById("useVadCheckbox");
   const languageDropdown = document.getElementById('languageDropdown');
   const taskDropdown = document.getElementById('taskDropdown');
   const modelSizeDropdown = document.getElementById('modelSizeDropdown');
@@ -31,6 +32,12 @@ document.addEventListener("DOMContentLoaded", function () {
     }
   });
 
+  chrome.storage.local.get("useVadState", ({ useVadState }) => {
+    if (useVadState !== undefined) {
+      useVadCheckbox.checked = useVadState;
+    }
+  });
+
   chrome.storage.local.get("selectedLanguage", ({ selectedLanguage: storedLanguage }) => {
     if (storedLanguage !== undefined) {
       languageDropdown.value = storedLanguage;
@@ -79,7 +86,8 @@ document.addEventListener("DOMContentLoaded", function () {
         port: port,
         language: selectedLanguage,
         task: selectedTask,
-        modelSize: selectedModelSize
+        modelSize: selectedModelSize,
+        useVad: useVadCheckbox.checked,
       }, () => {
         // Update capturing state in storage and toggle the buttons
         chrome.storage.local.set({ capturingState: { isCapturing: true } }, () => {
@@ -118,7 +126,8 @@ document.addEventListener("DOMContentLoaded", function () {
   function toggleCaptureButtons(isCapturing) {
     startButton.disabled = isCapturing;
     stopButton.disabled = !isCapturing;
-    useServerCheckbox.disabled = isCapturing; 
+    useServerCheckbox.disabled = isCapturing;
+    useVadCheckbox.disabled = isCapturing;
     modelSizeDropdown.disabled = isCapturing;
     languageDropdown.disabled = isCapturing;
     taskDropdown.disabled = isCapturing; 
@@ -132,6 +141,11 @@ document.addEventListener("DOMContentLoaded", function () {
     chrome.storage.local.set({ useServerState });
   });
 
+  useVadCheckbox.addEventListener("change", () => {
+    const useVadState = useVadCheckbox.checked;
+    chrome.storage.local.set({ useVadState });
+  });
+
   languageDropdown.addEventListener('change', function() {
     if (languageDropdown.value === "") {
       selectedLanguage = null;

Audio-Transcription-Firefox/content.js

@@ -74,7 +74,8 @@ function startRecording(data) {
             uid: uuid,
             language: data.language,
             task: data.task,
-            model: data.modelSize
+            model: data.modelSize,
+            use_vad: data.useVad
         })
       );
     };

Audio-Transcription-Firefox/popup.html

@@ -15,6 +15,10 @@
     <input type="checkbox" id="useServerCheckbox">
     <label for="useServerCheckbox">Use Collabora Whisper-Live Server</label>
   </div>
+  <div class="checkbox-container">
+    <input type="checkbox" id="useVadCheckbox">
+    <label for="useVadCheckbox">Use Voice Activity Detection</label>
+  </div>
   <textarea id="waitTextBox" style="display: none;"></textarea>
   <div class="dropdown-container">
     <label for="languageDropdown">Select Language:</label>

Audio-Transcription-Firefox/popup.js

@@ -3,6 +3,7 @@ document.addEventListener("DOMContentLoaded", function() {
   const stopButton = document.getElementById("stopCapture");
 
   const useServerCheckbox = document.getElementById("useServerCheckbox");
+  const useVadCheckbox = document.getElementById("useVadCheckbox");
   const languageDropdown = document.getElementById('languageDropdown');
   const taskDropdown = document.getElementById('taskDropdown');
   const modelSizeDropdown = document.getElementById('modelSizeDropdown');
@@ -34,6 +35,12 @@ document.addEventListener("DOMContentLoaded", function() {
     }
   });
 
+  browser.storage.local.get("useVadState", ({ useVadState }) => {
+    if (useVadState !== undefined) {
+      useVadCheckbox.checked = useVadState;
+    }
+  });
+
   browser.storage.local.get("selectedLanguage", ({ selectedLanguage: storedLanguage }) => {
     if (storedLanguage !== undefined) {
       languageDropdown.value = storedLanguage;
@@ -76,7 +83,8 @@ document.addEventListener("DOMContentLoaded", function() {
               port: port,
               language: selectedLanguage,
               task: selectedTask,
-              modelSize: selectedModelSize
+              modelSize: selectedModelSize,
+              useVad: useVadCheckbox.checked,
             } 
           });
         toggleCaptureButtons(true);
@@ -115,6 +123,7 @@ document.addEventListener("DOMContentLoaded", function() {
     startButton.disabled = isCapturing;
     stopButton.disabled = !isCapturing;
     useServerCheckbox.disabled = isCapturing;
+    useVadCheckbox.disabled = isCapturing;
     modelSizeDropdown.disabled = isCapturing;
     languageDropdown.disabled = isCapturing;
     taskDropdown.disabled = isCapturing; 
@@ -128,6 +137,11 @@ document.addEventListener("DOMContentLoaded", function() {
     browser.storage.local.set({ useServerState });
   });
 
+  useVadCheckbox.addEventListener("change", () => {
+    const useVadState = useVadCheckbox.checked;
+    browser.storage.local.set({ useVadState });
+  });
+
   languageDropdown.addEventListener('change', function() {
     if (languageDropdown.value === "") {
       selectedLanguage = null;

README.md

@@ -1,12 +1,18 @@
-# whisper-live
-A nearly-live implementation of OpenAI's Whisper.
+# WhisperLive
 
-This project is a real-time transcription application that uses the OpenAI Whisper model to convert speech input into text output. It can be used to transcribe both live audio input from microphone and pre-recorded audio files.
+<h2 align="center">
+  <a href="https://www.youtube.com/watch?v=0PHWCApIcCI"><img
+src="https://img.youtube.com/vi/0PHWCApIcCI/0.jpg" style="background-color:rgba(0,0,0,0);" height=300 alt="WhisperLive"></a>
+  <br><br>A nearly-live implementation of OpenAI's Whisper.
+<br><br>
+</h2>
 
-Unlike traditional speech recognition systems that rely on continuous audio streaming, we use [voice activity detection (VAD)](https://github.com/snakers4/silero-vad) to detect the presence of speech and only send the audio data to whisper when speech is detected. This helps to reduce the amount of data sent to the whisper model and improves the accuracy of the transcription output.
+This project is a real-time transcription application that uses the OpenAI Whisper model
+to convert speech input into text output. It can be used to transcribe both live audio
+input from microphone and pre-recorded audio files.
 
 ## Installation
-- Install PyAudio and ffmpeg
+- Install PyAudio
 ```bash
  bash scripts/setup.sh
 ```
@@ -30,7 +36,7 @@ python3 run_server.py --port 9090 \
   
 # running with custom model
 python3 run_server.py --port 9090 \
-                      --backend faster_whisper
+                      --backend faster_whisper \
                       -fw "/path/to/custom/faster/whisper/model"
 ```
 
@@ -47,10 +53,34 @@ python3 run_server.py -p 9090 \
                       -trt /home/TensorRT-LLM/examples/whisper/whisper_small \
                       -m
 ```
+#### Controlling OpenMP Threads
+To control the number of threads used by OpenMP, you can set the `OMP_NUM_THREADS` environment variable. This is useful for managing CPU resources and ensuring consistent performance. If not specified, `OMP_NUM_THREADS` is set to `1` by default. You can change this by using the `--omp_num_threads` argument:
+```bash
+python3 run_server.py --port 9090 \
+                      --backend faster_whisper \
+                      --omp_num_threads 4
+```
+
+#### Single model mode
+By default, when running the server without specifying a model, the server will instantiate a new whisper model for every client connection. This has the advantage, that the server can use different model sizes, based on the client's requested model size. On the other hand, it also means you have to wait for the model to be loaded upon client connection and you will have increased (V)RAM usage.
+
+When serving a custom TensorRT model using the `-trt` or a custom faster_whisper model using the `-fw` option, the server will instead only instantiate the custom model once and then reuse it for all client connections.
+
+If you don't want this, set `--no_single_model`.
 
 
 ### Running the Client
-- To transcribe an audio file:
+- Initializing the client with below parameters:
+  - `lang`: Language of the input audio, applicable only if using a multilingual model.
+  - `translate`: If set to `True` then translate from any language to `en`.
+  - `model`: Whisper model size.
+  - `use_vad`: Whether to use `Voice Activity Detection` on the server.
+  - `save_output_recording`: Set to True to save the microphone input as a `.wav` file during live transcription. This option is helpful for recording sessions for later playback or analysis. Defaults to `False`. 
+  - `output_recording_filename`: Specifies the `.wav` file path where the microphone input will be saved if `save_output_recording` is set to `True`.
+  - `max_clients`: Specifies the maximum number of clients the server should allow. Defaults to 4.
+  - `max_connection_time`: Maximum connection time for each client in seconds. Defaults to 600.
+  - `mute_audio_playback`: Whether to mute audio playback when transcribing an audio file. Defaults to False.
+
 ```python
 from whisper_live.client import TranscriptionClient
 client = TranscriptionClient(
@@ -58,58 +88,68 @@ client = TranscriptionClient(
   9090,
   lang="en",
   translate=False,
-  model="small"
+  model="small",                                      # also support hf_model => `Systran/faster-whisper-small`
+  use_vad=False,
+  save_output_recording=True,                         # Only used for microphone input, False by Default
+  output_recording_filename="./output_recording.wav", # Only used for microphone input
+  max_clients=4,
+  max_connection_time=600,
+  mute_audio_playback=False,                          # Only used for file input, False by Default
 )
+```
+It connects to the server running on localhost at port 9090. Using a multilingual model, language for the transcription will be automatically detected. You can also use the language option to specify the target language for the transcription, in this case, English ("en"). The translate option should be set to `True` if we want to translate from the source language to English and `False` if we want to transcribe in the source language.
 
+- Transcribe an audio file:
+```python
 client("tests/jfk.wav")
 ```
-This command transcribes the specified audio file (audio.wav) using the Whisper model. It connects to the server running on localhost at port 9090. Using a multilingual model, language for the transcription will be automatically detected. You can also use the language option to specify the target language for the transcription, in this case, English ("en"). The translate option should be set to `True` if we want to translate from the source language to English and `False` if we want to transcribe in the source language.
 
 - To transcribe from microphone:
 ```python
-from whisper_live.client import TranscriptionClient
-client = TranscriptionClient(
-  "localhost",
-  9090,
-  lang="hi",
-  translate=True,
-  model="small"
-)
 client()
 ```
-This command captures audio from the microphone and sends it to the server for transcription. It uses the multilingual model with `hi` as the selected language. We use whisper `small` by default but can be changed to any other option based on the requirements and the hardware running the server.
+
+- To transcribe from a RTSP stream:
+```python
+client(rtsp_url="rtsp://admin:admin@192.168.0.1/rtsp")
+```
 
 - To transcribe from a HLS stream:
 ```python
-from whisper_live.client import TranscriptionClient
-client = TranscriptionClient(host, port, lang="en", translate=False) 
-client(hls_url="http://as-hls-ww-live.akamaized.net/pool_904/live/ww/bbc_1xtra/bbc_1xtra.isml/bbc_1xtra-audio%3d96000.norewind.m3u8") 
+client(hls_url="http://as-hls-ww-live.akamaized.net/pool_904/live/ww/bbc_1xtra/bbc_1xtra.isml/bbc_1xtra-audio%3d96000.norewind.m3u8")
 ```
-This command streams audio into the server from a HLS stream. It uses the same options as the previous command, using the multilingual model and specifying the target language and task.
 
-## Transcribe audio from browser
-- Run the server with your desired backend as shown [here](https://github.com/collabora/WhisperLive?tab=readme-ov-file#running-the-server)
-
-### Chrome Extension
-- Refer to [Audio-Transcription-Chrome](https://github.com/collabora/whisper-live/tree/main/Audio-Transcription-Chrome#readme) to use Chrome extension.
-
-### Firefox Extension
-- Refer to [Audio-Transcription-Firefox](https://github.com/collabora/whisper-live/tree/main/Audio-Transcription-Firefox#readme) to use Mozilla Firefox extension.
+## Browser Extensions
+- Run the server with your desired backend as shown [here](https://github.com/collabora/WhisperLive?tab=readme-ov-file#running-the-server).
+- Transcribe audio directly from your browser using our Chrome or Firefox extensions. Refer to [Audio-Transcription-Chrome](https://github.com/collabora/whisper-live/tree/main/Audio-Transcription-Chrome#readme) and [Audio-Transcription-Firefox](https://github.com/collabora/whisper-live/tree/main/Audio-Transcription-Firefox#readme) for setup instructions.
 
 ## Whisper Live Server in Docker
 - GPU
   - Faster-Whisper
   ```bash
-  docker build . -t whisper-live -f docker/Dockerfile.gpu
-  docker run -it --gpus all -p 9090:9090 whisper-live:latest
+  docker run -it --gpus all -p 9090:9090 ghcr.io/collabora/whisperlive-gpu:latest
   ```
 
-  - TensorRT. Follow [TensorRT_whisper readme](https://github.com/collabora/WhisperLive/blob/main/TensorRT_whisper.md) in order to setup docker and use TensorRT backend. We provide a pre-built docker image which has TensorRT-LLM built and ready to use.
+  - TensorRT. 
+  ```bash
+  docker run -p 9090:9090 --runtime=nvidia --gpus all --entrypoint /bin/bash -it ghcr.io/collabora/whisperlive-tensorrt
+
+  # Build small.en engine
+  bash build_whisper_tensorrt.sh /app/TensorRT-LLM-examples small.en        # float16
+  bash build_whisper_tensorrt.sh /app/TensorRT-LLM-examples small.en int8   # int8 weight only quantization
+  bash build_whisper_tensorrt.sh /app/TensorRT-LLM-examples small.en int4   # int4 weight only quantization
+
+  # Run server with small.en
+  python3 run_server.py --port 9090 \
+                        --backend tensorrt \
+                        --trt_model_path "/app/TensorRT-LLM-examples/whisper/whisper_small_en_float16"
+                        --trt_model_path "/app/TensorRT-LLM-examples/whisper/whisper_small_en_int8"
+                        --trt_model_path "/app/TensorRT-LLM-examples/whisper/whisper_small_en_int4"
+  ```
 
 - CPU
 ```bash
-docker build . -t whisper-live -f docker/Dockerfile.cpu
-docker run -it -p 9090:9090 whisper-live:latest
+docker run -it -p 9090:9090 ghcr.io/collabora/whisperlive-cpu:latest
 ```
 **Note**: By default we use "small" model size. To build docker image for a different model size, change the size in server.py and then build the docker image.
 
@@ -140,6 +180,5 @@ We are available to help you with both Open Source and proprietary AI projects.
   publisher = {GitHub},
   journal = {GitHub repository},
   howpublished = {\url{https://github.com/snakers4/silero-vad}},
-  commit = {insert_some_commit_here},
   email = {hello@silero.ai}
 }

TensorRT_whisper.md

@@ -1,67 +1,38 @@
-# Whisper-TensorRT
+# WhisperLive-TensorRT
 We have only tested the TensorRT backend in docker so, we recommend docker for a smooth TensorRT backend setup.
-**Note**: We use [our fork to setup TensorRT](https://github.com/makaveli10/TensorRT-LLM)
+**Note**: We use `tensorrt_llm==0.15.0.dev2024111200`
 
 ## Installation
 - Install [docker](https://docs.docker.com/engine/install/)
 - Install [nvidia-container-toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html)
 
-- Clone this repo.
+- Run WhisperLive TensorRT in docker
 ```bash
-git clone https://github.com/collabora/WhisperLive.git
-cd WhisperLive
+docker run -p 9090:9090 --runtime=nvidia --gpus all --entrypoint /bin/bash -it ghcr.io/collabora/whisperlive-tensorrt:latest
 ```
 
-- Pull the TensorRT-LLM docker image which we prebuilt for WhisperLive TensorRT backend.
-```bash
-docker pull ghcr.io/collabora/whisperbot-base:latest
-```
-
-- Next, we run the docker image and mount WhisperLive repo to the containers `/home` directory.
-```bash
-docker run -it --gpus all --shm-size=8g \
-       --ipc=host --ulimit memlock=-1 --ulimit stack=67108864 \
-       -v /path/to/WhisperLive:/home/WhisperLive \
-       ghcr.io/collabora/whisperbot-base:latest
-```
-
-- Make sure to test the installation. 
-```bash
-# export ENV=${ENV:-/etc/shinit_v2} 
-# source $ENV
-python -c "import torch; import tensorrt; import tensorrt_llm"
-```
-**NOTE**: Uncomment and update library paths if imports fail.
-
 ## Whisper TensorRT Engine
-- We build `small.en` and `small` multilingual TensorRT engine. The script logs the path of the directory with Whisper TensorRT engine. We need the model_path to run the server.
+- We build `small.en` and `small` multilingual TensorRT engine as examples below. The script logs the path of the directory with Whisper TensorRT engine. We need that model_path to run the server.
 ```bash
 # convert small.en
-bash scripts/build_whisper_tensorrt.sh /root/TensorRT-LLM-examples small.en
+bash build_whisper_tensorrt.sh /app/TensorRT-LLM-examples small.en        # float16
+bash build_whisper_tensorrt.sh /app/TensorRT-LLM-examples small.en int8   # int8 weight only quantization
+bash build_whisper_tensorrt.sh /app/TensorRT-LLM-examples small.en int4   # int4 weight only quantization
 
 # convert small multilingual model
-bash scripts/build_whisper_tensorrt.sh /root/TensorRT-LLM-examples small
+bash build_whisper_tensorrt.sh /app/TensorRT-LLM-examples small
 ```
 
 ## Run WhisperLive Server with TensorRT Backend
 ```bash
-cd /home/WhisperLive
-
-# Install requirements
-bash scripts/setup.sh
-pip install -r requirements/server.txt
-
-# Required to create mel spectogram
-wget --directory-prefix=assets assets/mel_filters.npz https://raw.githubusercontent.com/openai/whisper/main/whisper/assets/mel_filters.npz
-
 # Run English only model
 python3 run_server.py --port 9090 \
                       --backend tensorrt \
-                      --trt_model_path "path/to/whisper_trt/from/build/step"
+                      --trt_model_path "/app/TensorRT-LLM-examples/whisper/whisper_small_en_float16"
 
 # Run Multilingual model
 python3 run_server.py --port 9090 \
                       --backend tensorrt \
-                      --trt_model_path "path/to/whisper_trt/from/build/step" \
+                      --trt_model_path "/app/TensorRT-LLM-examples/whisper/whisper_small_float16" \
                       --trt_multilingual
 ```

docker/Dockerfile.cpu

@@ -1,45 +1,25 @@
-FROM ubuntu:focal
+FROM python:3.10-bookworm
 
 ARG DEBIAN_FRONTEND=noninteractive
 
-# Remove any third-party apt sources to avoid issues with expiring keys.
-RUN rm -f /etc/apt/sources.list.d/*.list
+# install lib required for pyaudio
+RUN apt update && apt install -y portaudio19-dev && apt-get clean && rm -rf /var/lib/apt/lists/*
 
-# Install some basic utilities.
-RUN apt-get update && apt-get install -y \
-    curl \
-    ca-certificates \
-    sudo \
-    git \
-    bzip2 \
-    libx11-6 \
- && rm -rf /var/lib/apt/lists/*
+# update pip to support for whl.metadata -> less downloading
+RUN pip install --no-cache-dir -U "pip>=24"
 
-RUN apt update
-
-# install python
-RUN apt install software-properties-common -y && \
-    add-apt-repository ppa:deadsnakes/ppa && \
-    apt update
-
-RUN apt install python3-dev -y && \
-    apt install python-is-python3
-    
-
-# install pip
-RUN apt install python3-pip -y
-
-# Create a working directory.
+# create a working directory
 RUN mkdir /app
 WORKDIR /app
 
-COPY scripts/setup.sh /app
-COPY requirements/ /app
+# install pytorch, but without the nvidia-libs that are only necessary for gpu
+RUN pip install --no-cache-dir torch --index-url https://download.pytorch.org/whl/cpu
 
-RUN bash setup.sh
-RUN pip install -r server.txt
+# install the requirements for running the whisper-live server
+COPY requirements/server.txt /app/
+RUN pip install --no-cache-dir -r server.txt && rm server.txt
 
 COPY whisper_live /app/whisper_live
-
 COPY run_server.py /app
+
 CMD ["python", "run_server.py"]

docker/Dockerfile.gpu

@@ -1,47 +1,26 @@
-FROM nvidia/cuda:11.2.2-cudnn8-runtime-ubuntu20.04
+FROM python:3.10-bookworm
 
 ARG DEBIAN_FRONTEND=noninteractive
 
-# Remove any third-party apt sources to avoid issues with expiring keys.
-RUN rm -f /etc/apt/sources.list.d/*.list
+# install lib required for pyaudio
+RUN apt update && apt install -y portaudio19-dev && apt-get clean && rm -rf /var/lib/apt/lists/*
 
-# Install some basic utilities.
-RUN apt-get update && apt-get install -y \
-    curl \
-    ca-certificates \
-    sudo \
-    git \
-    bzip2 \
-    libx11-6 \
- && rm -rf /var/lib/apt/lists/*
+# update pip to support for whl.metadata -> less downloading
+RUN pip install --no-cache-dir -U "pip>=24"
 
-RUN apt update
-
-# install python
-RUN apt install software-properties-common -y && \
-    add-apt-repository ppa:deadsnakes/ppa && \
-    apt update
-
-RUN apt install python3-dev -y && \
-    apt install python-is-python3
-    
-
-# install pip
-RUN apt install python3-pip -y
-
-# Create a working directory.
+# create a working directory
 RUN mkdir /app
 WORKDIR /app
 
-COPY scripts/setup.sh /app
-COPY requirements/ /app
+# install the requirements for running the whisper-live server
+COPY requirements/server.txt /app/
+RUN pip install --no-cache-dir -r server.txt && rm server.txt
 
-RUN apt update --fix-missing
-RUN bash setup.sh
-RUN pip install -r server.txt
+# make the paths of the nvidia libs installed as wheels visible. equivalent to:
+# export LD_LIBRARY_PATH=`python3 -c 'import os; import nvidia.cublas.lib; import nvidia.cudnn.lib; print(os.path.dirname(nvidia.cublas.lib.__file__) + ":" + os.path.dirname(nvidia.cudnn.lib.__file__))'`
+ENV LD_LIBRARY_PATH="/usr/local/lib/python3.10/site-packages/nvidia/cublas/lib:/usr/local/lib/python3.10/site-packages/nvidia/cudnn/lib"
 
 COPY whisper_live /app/whisper_live
-
 COPY run_server.py /app
 
 CMD ["python", "run_server.py"]

docker/Dockerfile.tensorrt

@@ -0,0 +1,31 @@
+FROM nvidia/cuda:12.4.1-base-ubuntu22.04 AS base
+
+ARG DEBIAN_FRONTEND=noninteractive
+
+RUN apt-get update && apt-get install -y \
+    python3.10 python3-pip openmpi-bin libopenmpi-dev git git-lfs wget \
+    && rm -rf /var/lib/apt/lists/*
+
+FROM base AS devel
+RUN pip3 install --no-cache-dir -U tensorrt_llm==0.15.0.dev2024111200 --extra-index-url https://pypi.nvidia.com
+WORKDIR /app
+RUN git clone https://github.com/NVIDIA/TensorRT-LLM.git && cd TensorRT-LLM && \
+    git checkout c629546ce429623c8a163633095230154a6f0574 && cd ../ && \
+    mv TensorRT-LLM/examples ./TensorRT-LLM-examples && \
+    rm -rf TensorRT-LLM
+
+
+FROM devel AS release
+WORKDIR /app
+COPY assets/ ./assets
+RUN wget -nc -P assets/ https://raw.githubusercontent.com/openai/whisper/main/whisper/assets/mel_filters.npz
+
+COPY scripts/setup.sh ./
+RUN apt update && bash setup.sh && rm setup.sh
+
+COPY requirements/server.txt .
+RUN pip install --no-cache-dir -r server.txt && rm server.txt
+RUN pip install pynvml==11.5.0
+COPY whisper_live ./whisper_live
+COPY scripts/build_whisper_tensorrt.sh .
+COPY run_server.py .

requirements/client.txt

@@ -1,4 +1,4 @@
 PyAudio
-ffmpeg-python
+av
 scipy
 websocket-client

requirements/server.txt

@@ -1,10 +1,13 @@
-faster-whisper==0.10.0
-torch
+faster-whisper==1.1.0
 websockets
-onnxruntime==1.16.0
+onnxruntime==1.17.0
 numba
-openai-whisper
 kaldialign
 soundfile
-ffmpeg-python
 scipy
+av
+jiwer
+evaluate
+numpy<2
+openai-whisper==20240930
+tokenizers==0.20.3

run_server.py

@@ -1,18 +1,18 @@
 import argparse
-from whisper_live.server import TranscriptionServer
+import os
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser()
     parser.add_argument('--port', '-p',
-                        type=int, 
+                        type=int,
                         default=9090,
                         help="Websocket port to run the server on.")
     parser.add_argument('--backend', '-b',
-                        type=str, 
-                        default='faster_whisper', 
+                        type=str,
+                        default='faster_whisper',
                         help='Backends from ["tensorrt", "faster_whisper"]')
     parser.add_argument('--faster_whisper_custom_model_path', '-fw',
-                        type=str, default=None, 
+                        type=str, default=None,
                         help="Custom Faster Whisper Model")
     parser.add_argument('--trt_model_path', '-trt',
                         type=str,
@@ -21,18 +21,30 @@ if __name__ == "__main__":
     parser.add_argument('--trt_multilingual', '-m',
                         action="store_true",
                         help='Boolean only for TensorRT model. True if multilingual.')
+    parser.add_argument('--omp_num_threads', '-omp',
+                        type=int,
+                        default=1,
+                        help="Number of threads to use for OpenMP")
+    parser.add_argument('--no_single_model', '-nsm',
+                        action='store_true',
+                        help='Set this if every connection should instantiate its own model. Only relevant for custom model, passed using -trt or -fw.')
     args = parser.parse_args()
 
     if args.backend == "tensorrt":
         if args.trt_model_path is None:
             raise ValueError("Please Provide a valid tensorrt model path")
 
+    if "OMP_NUM_THREADS" not in os.environ:
+        os.environ["OMP_NUM_THREADS"] = str(args.omp_num_threads)
+
+    from whisper_live.server import TranscriptionServer
     server = TranscriptionServer()
     server.run(
         "0.0.0.0",
-        port=args.port, 
+        port=args.port,
         backend=args.backend,
         faster_whisper_custom_model_path=args.faster_whisper_custom_model_path,
         whisper_tensorrt_path=args.trt_model_path,
-        trt_multilingual=args.trt_multilingual
+        trt_multilingual=args.trt_multilingual,
+        single_model=not args.no_single_model,
     )

scripts/build_whisper_tensorrt.sh

@@ -38,12 +38,24 @@ download_and_build_model() {
         "large-v3" | "large")
             model_url="https://openaipublic.azureedge.net/main/whisper/models/e5b1a55b89c1367dacf97e3e19bfd829a01529dbfdeefa8caeb59b3f1b81dadb/large-v3.pt"
             ;;
+        "large-v3-turbo" | "turbo")
+            model_url="https://openaipublic.azureedge.net/main/whisper/models/aff26ae408abcba5fbf8813c21e62b0941638c5f6eebfb145be0c9839262a19a/large-v3-turbo.pt"
+            ;;
         *)
             echo "Invalid model name: $model_name"
             exit 1
             ;;
     esac
 
+    if [ "$model_name" == "turbo" ]; then
+        model_name="large-v3-turbo"
+    fi
+
+    local inference_precision="float16"
+    local weight_only_precision="${2:-float16}"
+    local max_beam_width=4
+    local max_batch_size=1
+
     echo "Downloading $model_name..."
     # wget --directory-prefix=assets "$model_url"
     # echo "Download completed: ${model_name}.pt"
@@ -54,11 +66,43 @@ download_and_build_model() {
         echo "${model_name}.pt already exists in assets directory."
     fi
 
-    local output_dir="whisper_${model_name//./_}"
+    local sanitized_model_name="${model_name//./_}"
+    local checkpoint_dir="whisper_${sanitized_model_name}_weights_${weight_only_precision}"
+    local output_dir="whisper_${sanitized_model_name}_${weight_only_precision}"
     echo "$output_dir"
-    echo "Running build script for $model_name with output directory $output_dir"
-    python3 build.py --output_dir "$output_dir" --use_gpt_attention_plugin --use_gemm_plugin --use_bert_attention_plugin --model_name "$model_name"
-    echo "Whisper $model_name TensorRT engine built."
+    echo "Converting model weights for $model_name..."
+    python3 convert_checkpoint.py \
+        $( [[ "$weight_only_precision" == "int8" || "$weight_only_precision" == "int4" ]] && echo "--use_weight_only --weight_only_precision $weight_only_precision" ) \
+        --output_dir "$checkpoint_dir" --model_name "$model_name"
+    
+    echo "Building encoder for $model_name..."
+    trtllm-build \
+        --checkpoint_dir "${checkpoint_dir}/encoder" \
+        --output_dir "${output_dir}/encoder" \
+        --moe_plugin disable \
+        --enable_xqa disable \
+        --max_batch_size "$max_batch_size" \
+        --gemm_plugin disable \
+        --bert_attention_plugin "$inference_precision" \
+        --max_input_len 3000 \
+        --max_seq_len 3000
+    
+    echo "Building decoder for $model_name..."
+    trtllm-build \
+        --checkpoint_dir "${checkpoint_dir}/decoder" \
+        --output_dir "${output_dir}/decoder" \
+        --moe_plugin disable \
+        --enable_xqa disable \
+        --max_beam_width "$max_beam_width" \
+        --max_batch_size "$max_batch_size" \
+        --max_seq_len 200 \
+        --max_input_len 14 \
+        --max_encoder_input_len 3000 \
+        --gemm_plugin "$inference_precision" \
+        --bert_attention_plugin "$inference_precision" \
+        --gpt_attention_plugin "$inference_precision"
+
+    echo "TensorRT LLM engine built for $model_name."
     echo "========================================="
     echo "Model is located at: $(pwd)/$output_dir"
 }
@@ -70,8 +114,9 @@ fi
 
 tensorrt_examples_dir="$1"
 model_name="${2:-small.en}"
+weight_only_precision="${3:-float16}"  # Default to float16 if not provided
 
-cd $1/whisper
+cd $tensorrt_examples_dir/whisper
 pip install --no-deps -r requirements.txt
 
-download_and_build_model "$model_name"
+download_and_build_model "$model_name" "$weight_only_precision"

scripts/setup.sh

@@ -1,3 +1,3 @@
 #! /bin/bash
 
-apt-get install portaudio19-dev ffmpeg wget -y
+apt-get install portaudio19-dev wget -y

setup.py

@@ -10,49 +10,51 @@ HERE = pathlib.Path(__file__).parent
 README = (HERE / "README.md").read_text()
 
 # This call to setup() does all the work
-setup(name="whisper-live",
-      version=__version__,
-      description="A nearly-live implementation of OpenAI's Whisper.",
-      long_description=README,
-      long_description_content_type="text/markdown",
-      include_package_data=True,
-      url="https://github.com/collabora/WhisperLive",
-      author="Collabora Ltd",
-      author_email="vineet.suryan@collabora.com",
-      license="MIT",
-      classifiers=[
-          "Development Status :: 4 - Beta",
-          "Intended Audience :: Developers",
-          "Intended Audience :: Science/Research",
-          "License :: OSI Approved :: MIT License",
-          "Programming Language :: Python :: 3",
-          "Programming Language :: Python :: 3 :: Only",
-          "Programming Language :: Python :: 3.8",
-          "Programming Language :: Python :: 3.9",
-          "Topic :: Scientific/Engineering :: Artificial Intelligence",
-      ],
-      packages=find_packages(
-        exclude=("examples",
-                 "Audio-Transcription-Chrome",
-                 "Audio-Transcription-Firefox",
-                 "requirements",
-                 "whisper-finetuning"
-                )
-      ),
-      install_requires=[
+setup(
+    name="whisper_live",
+    version=__version__,
+    description="A nearly-live implementation of OpenAI's Whisper.",
+    long_description=README,
+    long_description_content_type="text/markdown",
+    include_package_data=True,
+    url="https://github.com/collabora/WhisperLive",
+    author="Collabora Ltd",
+    author_email="vineet.suryan@collabora.com",
+    license="MIT",
+    classifiers=[
+        "Development Status :: 4 - Beta",
+        "Intended Audience :: Developers",
+        "Intended Audience :: Science/Research",
+        "License :: OSI Approved :: MIT License",
+        "Programming Language :: Python :: 3",
+        "Programming Language :: Python :: 3 :: Only",
+        "Programming Language :: Python :: 3.8",
+        "Programming Language :: Python :: 3.9",
+        "Topic :: Scientific/Engineering :: Artificial Intelligence",
+    ],
+    packages=find_packages(
+        exclude=(
+            "examples",
+            "Audio-Transcription-Chrome",
+            "Audio-Transcription-Firefox",
+            "requirements",
+            "whisper-finetuning"
+        )
+    ),
+    install_requires=[
         "PyAudio",
-        "faster-whisper==0.10.0",
+        "faster-whisper==1.1.0",
         "torch",
         "torchaudio",
         "websockets",
         "onnxruntime==1.16.0",
-        "ffmpeg-python",
         "scipy",
         "websocket-client",
         "numba",
-        "openai-whisper",
+        "openai-whisper==20240930",
         "kaldialign",
         "soundfile",
-      ],
-      python_requires=">=3.8"
+        "tokenizers==0.20.3"
+    ],
+    python_requires=">=3.8"
 )

tests/test_client.py

@@ -0,0 +1,158 @@
+import json
+import os
+import scipy
+import websocket
+import copy
+import unittest
+from unittest.mock import patch, MagicMock
+from whisper_live.client import Client, TranscriptionClient, TranscriptionTeeClient
+from whisper_live.utils import resample
+from pathlib import Path
+
+
+class BaseTestCase(unittest.TestCase):
+    @patch('whisper_live.client.websocket.WebSocketApp')
+    @patch('whisper_live.client.pyaudio.PyAudio')
+    def setUp(self, mock_pyaudio, mock_websocket):
+        self.mock_pyaudio_instance = MagicMock()
+        mock_pyaudio.return_value = self.mock_pyaudio_instance
+        self.mock_stream = MagicMock()
+        self.mock_pyaudio_instance.open.return_value = self.mock_stream
+
+        self.mock_ws_app = mock_websocket.return_value
+        self.mock_ws_app.send = MagicMock()
+
+        self.client = TranscriptionClient(host='localhost', port=9090, lang="en").client
+
+        self.mock_pyaudio = mock_pyaudio
+        self.mock_websocket = mock_websocket
+        self.mock_audio_packet = b'\x00\x01\x02\x03'
+
+    def tearDown(self):
+        self.client.close_websocket()
+        self.mock_pyaudio.stop()
+        self.mock_websocket.stop()
+        del self.client
+
+class TestClientWebSocketCommunication(BaseTestCase):
+    def test_websocket_communication(self):
+        expected_url = 'ws://localhost:9090'
+        self.mock_websocket.assert_called()
+        self.assertEqual(self.mock_websocket.call_args[0][0], expected_url)
+
+
+class TestClientCallbacks(BaseTestCase):
+    def test_on_open(self):
+        expected_message = json.dumps({
+            "uid": self.client.uid,
+            "language": self.client.language,
+            "task": self.client.task,
+            "model": self.client.model,
+            "use_vad": True,
+            "max_clients": 4,
+            "max_connection_time": 600,
+        })
+        self.client.on_open(self.mock_ws_app)
+        self.mock_ws_app.send.assert_called_with(expected_message)
+
+    def test_on_message(self):
+        message = json.dumps(
+            {
+                "uid": self.client.uid,
+                "message": "SERVER_READY",
+                "backend": "faster_whisper"
+            }
+        )
+        self.client.on_message(self.mock_ws_app, message)
+
+        message = json.dumps({
+            "uid": self.client.uid,
+            "segments": [
+                {"start": 0, "end": 1, "text": "Test transcript", "completed": True},
+                {"start": 1, "end": 2, "text": "Test transcript 2", "completed": True},
+                {"start": 2, "end": 3, "text": "Test transcript 3", "completed": True}
+            ]
+        })
+        self.client.on_message(self.mock_ws_app, message)
+
+        # Assert that the transcript was updated correctly
+        self.assertEqual(len(self.client.transcript), 3)
+        self.assertEqual(self.client.transcript[1]['text'], "Test transcript 2")
+
+    def test_on_close(self):
+        close_status_code = 1000
+        close_msg = "Normal closure"
+        self.client.on_close(self.mock_ws_app, close_status_code, close_msg)
+
+        self.assertFalse(self.client.recording)
+        self.assertFalse(self.client.server_error)
+        self.assertFalse(self.client.waiting)
+
+    def test_on_error(self):
+        error_message = "Test Error"
+        self.client.on_error(self.mock_ws_app, error_message)
+
+        self.assertTrue(self.client.server_error)
+        self.assertEqual(self.client.error_message, error_message)
+
+
+class TestAudioResampling(unittest.TestCase):
+    def test_resample_audio(self):
+        original_audio = "assets/jfk.flac"
+        expected_sr = 16000
+        resampled_audio = resample(original_audio, expected_sr)
+
+        sr, _ = scipy.io.wavfile.read(resampled_audio)
+        self.assertEqual(sr, expected_sr)
+
+        os.remove(resampled_audio)
+
+
+class TestSendingAudioPacket(BaseTestCase):
+    def test_send_packet(self):
+        self.client.send_packet_to_server(self.mock_audio_packet)
+        self.client.client_socket.send.assert_called_with(self.mock_audio_packet, websocket.ABNF.OPCODE_BINARY)
+
+class TestTee(BaseTestCase):
+    @patch('whisper_live.client.websocket.WebSocketApp')
+    @patch('whisper_live.client.pyaudio.PyAudio')
+    def setUp(self, mock_audio, mock_websocket):
+        super().setUp()
+        self.client2 = Client(host='localhost', port=9090, lang="es", translate=False, srt_file_path="transcript.srt")
+        self.client3 = Client(host='localhost', port=9090, lang="es", translate=True, srt_file_path="translation.srt")
+        # need a separate mock for each websocket
+        self.client3.client_socket = copy.deepcopy(self.client3.client_socket)
+        self.tee = TranscriptionTeeClient([self.client2, self.client3])
+
+    def tearDown(self):
+        self.tee.close_all_clients()
+        del self.tee
+        super().tearDown()
+
+    def test_invalid_constructor(self):
+        with self.assertRaises(Exception) as context:
+            TranscriptionTeeClient([])
+
+    def test_multicast_unconditional(self):
+        self.tee.multicast_packet(self.mock_audio_packet, True)
+        for client in self.tee.clients:
+            client.client_socket.send.assert_called_with(self.mock_audio_packet, websocket.ABNF.OPCODE_BINARY)
+
+    def test_multicast_conditional(self):
+        self.client2.recording = False
+        self.client3.recording = True
+        self.tee.multicast_packet(self.mock_audio_packet, False)
+        self.client2.client_socket.send.assert_not_called()
+        self.client3.client_socket.send.assert_called_with(self.mock_audio_packet, websocket.ABNF.OPCODE_BINARY)
+
+    def test_close_all(self):
+        self.tee.close_all_clients()
+        for client in self.tee.clients:
+            client.client_socket.close.assert_called()
+
+    def test_write_all_srt(self):
+        for client in self.tee.clients:
+            client.server_backend = "faster_whisper"
+        self.tee.write_all_clients_srt()
+        self.assertTrue(Path("transcript.srt").is_file())
+        self.assertTrue(Path("translation.srt").is_file())

tests/test_server.py

@@ -0,0 +1,148 @@
+import subprocess
+import time
+import json
+import unittest
+from unittest import mock
+
+import numpy as np
+import jiwer
+
+from websockets.exceptions import ConnectionClosed
+from whisper_live.server import TranscriptionServer, BackendType, ClientManager
+from whisper_live.client import Client, TranscriptionClient, TranscriptionTeeClient
+from whisper.normalizers import EnglishTextNormalizer
+
+
+class TestTranscriptionServerInitialization(unittest.TestCase):
+    def test_initialization(self):
+        server = TranscriptionServer()
+        server.client_manager = ClientManager(max_clients=4, max_connection_time=600)
+        self.assertEqual(server.client_manager.max_clients, 4)
+        self.assertEqual(server.client_manager.max_connection_time, 600)
+        self.assertDictEqual(server.client_manager.clients, {})
+        self.assertDictEqual(server.client_manager.start_times, {})
+
+
+class TestGetWaitTime(unittest.TestCase):
+    def setUp(self):
+        self.server = TranscriptionServer()
+        self.server.client_manager = ClientManager(max_clients=4, max_connection_time=600)
+        self.server.client_manager.start_times = {
+            'client1': time.time() - 120,
+            'client2': time.time() - 300
+        }
+        self.server.client_manager.max_connection_time = 600
+
+    def test_get_wait_time(self):
+        expected_wait_time = (600 - (time.time() - self.server.client_manager.start_times['client2'])) / 60
+        print(self.server.client_manager.get_wait_time(), expected_wait_time)
+        self.assertAlmostEqual(self.server.client_manager.get_wait_time(), expected_wait_time, places=2)
+
+
+class TestServerConnection(unittest.TestCase):
+    def setUp(self):
+        self.server = TranscriptionServer()
+
+    @mock.patch('websockets.WebSocketCommonProtocol')
+    def test_connection(self, mock_websocket):
+        mock_websocket.recv.return_value = json.dumps({
+            'uid': 'test_client',
+            'language': 'en',
+            'task': 'transcribe',
+            'model': 'tiny.en'
+        })
+        self.server.recv_audio(mock_websocket, BackendType("faster_whisper"))
+
+    @mock.patch('websockets.WebSocketCommonProtocol')
+    def test_recv_audio_exception_handling(self, mock_websocket):
+        mock_websocket.recv.side_effect = [json.dumps({
+            'uid': 'test_client',
+            'language': 'en',
+            'task': 'transcribe',
+            'model': 'tiny.en'
+        }),  np.array([1, 2, 3]).tobytes()]
+
+        with self.assertLogs(level="ERROR"):
+            self.server.recv_audio(mock_websocket, BackendType("faster_whisper"))
+
+        self.assertNotIn(mock_websocket, self.server.client_manager.clients)
+
+
+class TestServerInferenceAccuracy(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls):
+        cls.mock_pyaudio_patch = mock.patch('pyaudio.PyAudio')
+        cls.mock_pyaudio = cls.mock_pyaudio_patch.start()
+        cls.mock_pyaudio.return_value.open.return_value = mock.MagicMock()
+        
+        cls.server_process = subprocess.Popen(["python", "run_server.py"])
+        time.sleep(2)
+
+    @classmethod
+    def tearDownClass(cls):
+        cls.server_process.terminate()
+        cls.server_process.wait()
+
+    def setUp(self):
+        self.normalizer = EnglishTextNormalizer()
+
+    def check_prediction(self, srt_path):
+        gt = "And so my fellow Americans, ask not, what your country can do for you. Ask what you can do for your country!"
+        with open(srt_path, "r") as f:
+            lines = f.readlines()
+            prediction = " ".join([line.strip() for line in lines[2::4]])
+        prediction_normalized = self.normalizer(prediction)
+        gt_normalized = self.normalizer(gt)
+
+        # calculate WER
+        wer_score = jiwer.wer(gt_normalized, prediction_normalized)
+        self.assertLess(wer_score, 0.05)
+
+    def test_inference(self):
+        client = TranscriptionClient(
+            "localhost", "9090", model="base.en", lang="en",
+        )
+        client("assets/jfk.flac")
+        self.check_prediction("output.srt")
+
+    def test_simultaneous_inference(self):
+        client1 = Client(
+            "localhost", "9090", model="base.en", lang="en", srt_file_path="transcript1.srt")
+        client2 = Client(
+            "localhost", "9090", model="base.en", lang="en", srt_file_path="transcript2.srt")
+        tee = TranscriptionTeeClient([client1, client2])
+        tee("assets/jfk.flac")
+        self.check_prediction("transcript1.srt")
+        self.check_prediction("transcript2.srt")
+
+
+class TestExceptionHandling(unittest.TestCase):
+    def setUp(self):
+        self.server = TranscriptionServer()
+
+    @mock.patch('websockets.WebSocketCommonProtocol')
+    def test_connection_closed_exception(self, mock_websocket):
+        mock_websocket.recv.side_effect = ConnectionClosed(1001, "testing connection closed", rcvd_then_sent=mock.Mock())
+
+        with self.assertLogs(level="INFO") as log:
+            self.server.recv_audio(mock_websocket, BackendType("faster_whisper"))
+            self.assertTrue(any("Connection closed by client" in message for message in log.output))
+
+    @mock.patch('websockets.WebSocketCommonProtocol')
+    def test_json_decode_exception(self, mock_websocket):
+        mock_websocket.recv.return_value = "invalid json"
+
+        with self.assertLogs(level="ERROR") as log:
+            self.server.recv_audio(mock_websocket, BackendType("faster_whisper"))
+            self.assertTrue(any("Failed to decode JSON from client" in message for message in log.output))
+
+    @mock.patch('websockets.WebSocketCommonProtocol')
+    def test_unexpected_exception_handling(self, mock_websocket):
+        mock_websocket.recv.side_effect = RuntimeError("Unexpected error")
+
+        with self.assertLogs(level="ERROR") as log:
+            self.server.recv_audio(mock_websocket, BackendType("faster_whisper"))
+            for message in log.output:
+                print(message)
+            print()
+            self.assertTrue(any("Unexpected error" in message for message in log.output))

tests/test_vad.py

@@ -0,0 +1,26 @@
+import unittest
+import numpy as np
+from whisper_live.tensorrt_utils import load_audio
+from whisper_live.vad import VoiceActivityDetector
+
+
+class TestVoiceActivityDetection(unittest.TestCase):
+    def setUp(self):
+        self.vad = VoiceActivityDetector()
+        self.sample_rate = 16000
+
+    def generate_silence(self, duration_seconds):
+        return np.zeros(int(self.sample_rate * duration_seconds), dtype=np.float32)
+
+    def load_speech_segment(self, filepath):
+        return load_audio(filepath)
+
+    def test_vad_silence_detection(self):
+        silence = self.generate_silence(3)
+        is_speech_present = self.vad(silence.copy())
+        self.assertFalse(is_speech_present, "VAD incorrectly identified silence as speech.")
+
+    def test_vad_speech_detection(self):
+        audio_tensor = load_audio("assets/jfk.flac")
+        is_speech_present = self.vad(audio_tensor)
+        self.assertTrue(is_speech_present, "VAD failed to identify speech segment.")

whisper_live/__version__.py

@@ -1 +1 @@
-__version__="0.1.0"
+__version__ = "0.6.2"

whisper_live/client.py

@@ -1,76 +1,25 @@
 import os
+import shutil
 import wave
 
+import logging
 import numpy as np
-import scipy
-import ffmpeg
 import pyaudio
 import threading
-import textwrap
 import json
 import websocket
 import uuid
 import time
-
-
-def format_time(s):
-    """Convert seconds (float) to SRT time format."""
-    hours = int(s // 3600)
-    minutes = int((s % 3600) // 60)
-    seconds = int(s % 60)
-    milliseconds = int((s - int(s)) * 1000)
-    return f"{hours:02}:{minutes:02}:{seconds:02},{milliseconds:03}"
-
-def create_srt_file(segments, output_file):
-    with open(output_file, 'w', encoding='utf-8') as srt_file:
-        segment_number = 1
-        for segment in segments:
-            start_time = format_time(float(segment['start']))
-            end_time = format_time(float(segment['end']))
-            text = segment['text']
-            
-            srt_file.write(f"{segment_number}\n")
-            srt_file.write(f"{start_time} --> {end_time}\n")
-            srt_file.write(f"{text}\n\n")
-            
-            segment_number += 1
-
-
-def resample(file: str, sr: int = 16000):
-    """
-    # https://github.com/openai/whisper/blob/7858aa9c08d98f75575035ecd6481f462d66ca27/whisper/audio.py#L22
-    Open an audio file and read as mono waveform, resampling as necessary,
-    save the resampled audio
-
-    Args:
-        file (str): The audio file to open
-        sr (int): The sample rate to resample the audio if necessary
-    
-    Returns:
-        resampled_file (str): The resampled audio file
-    """
-    try:
-        # This launches a subprocess to decode audio while down-mixing and resampling as necessary.
-        # Requires the ffmpeg CLI and `ffmpeg-python` package to be installed.
-        out, _ = (
-            ffmpeg.input(file, threads=0)
-            .output("-", format="s16le", acodec="pcm_s16le", ac=1, ar=sr)
-            .run(cmd=["ffmpeg", "-nostdin"], capture_stdout=True, capture_stderr=True)
-        )
-    except ffmpeg.Error as e:
-        raise RuntimeError(f"Failed to load audio: {e.stderr.decode()}") from e
-    np_buffer = np.frombuffer(out, dtype=np.int16)
-
-    resampled_file = f"{file.split('.')[0]}_resampled.wav"
-    scipy.io.wavfile.write(resampled_file, sr, np_buffer.astype(np.int16))
-    return resampled_file
+import av
+import whisper_live.utils as utils
 
 
 class Client:
     """
-    Handles audio recording, streaming, and communication with a server using WebSocket.
+    Handles communication with a server using WebSocket.
     """
     INSTANCES = {}
+    END_OF_AUDIO = "END_OF_AUDIO"
 
     def __init__(
         self,
@@ -79,7 +28,11 @@ class Client:
         lang=None,
         translate=False,
         model="small",
-        srt_file_path="output.srt"
+        srt_file_path="output.srt",
+        use_vad=True,
+        log_transcription=True,
+        max_clients=4,
+        max_connection_time=600,
     ):
         """
         Initializes a Client instance for audio recording and streaming to a server.
@@ -93,36 +46,34 @@ class Client:
             port (int): The port number for the WebSocket server.
             lang (str, optional): The selected language for transcription. Default is None.
             translate (bool, optional): Specifies if the task is translation. Default is False.
+            model (str, optional): The whisper model to use (e.g., "small", "medium", "large"). Default is "small".
+            srt_file_path (str, optional): The file path to save the output SRT file. Default is "output.srt".
+            use_vad (bool, optional): Whether to enable voice activity detection. Default is True.
+            log_transcription (bool, optional): Whether to log transcription output to the console. Default is True.
+            max_clients (int, optional): Maximum number of client connections allowed. Default is 4.
+            max_connection_time (int, optional): Maximum allowed connection time in seconds. Default is 600.
         """
-        self.chunk = 4096
-        self.format = pyaudio.paInt16
-        self.channels = 1
-        self.rate = 16000
-        self.record_seconds = 60000
         self.recording = False
         self.task = "transcribe"
         self.uid = str(uuid.uuid4())
         self.waiting = False
-        self.last_response_recieved = None
+        self.last_response_received = None
         self.disconnect_if_no_response_for = 15
         self.language = lang
         self.model = model
         self.server_error = False
         self.srt_file_path = srt_file_path
+        self.use_vad = use_vad
+        self.last_segment = None
+        self.last_received_segment = None
+        self.log_transcription = log_transcription
+        self.max_clients = max_clients
+        self.max_connection_time = max_connection_time
 
         if translate:
             self.task = "translate"
 
-        self.timestamp_offset = 0.0
         self.audio_bytes = None
-        self.p = pyaudio.PyAudio()
-        self.stream = self.p.open(
-            format=self.format,
-            channels=self.channels,
-            rate=self.rate,
-            input=True,
-            frames_per_buffer=self.chunk,
-        )
 
         if host is not None and port is not None:
             socket_url = f"ws://{host}:{port}"
@@ -146,14 +97,48 @@ class Client:
         self.ws_thread.setDaemon(True)
         self.ws_thread.start()
 
-        self.frames = b""
         self.transcript = []
         print("[INFO]: * recording")
 
+    def handle_status_messages(self, message_data):
+        """Handles server status messages."""
+        status = message_data["status"]
+        if status == "WAIT":
+            self.waiting = True
+            print(f"[INFO]: Server is full. Estimated wait time {round(message_data['message'])} minutes.")
+        elif status == "ERROR":
+            print(f"Message from Server: {message_data['message']}")
+            self.server_error = True
+        elif status == "WARNING":
+            print(f"Message from Server: {message_data['message']}")
+
+    def process_segments(self, segments):
+        """Processes transcript segments."""
+        text = []
+        for i, seg in enumerate(segments):
+            if not text or text[-1] != seg["text"]:
+                text.append(seg["text"])
+                if i == len(segments) - 1 and not seg.get("completed", False):
+                    self.last_segment = seg
+                elif (self.server_backend == "faster_whisper" and seg.get("completed", False) and
+                      (not self.transcript or
+                        float(seg['start']) >= float(self.transcript[-1]['end']))):
+                    self.transcript.append(seg)
+        # update last received segment and last valid response time
+        if self.last_received_segment is None or self.last_received_segment != segments[-1]["text"]:
+            self.last_response_received = time.time()
+            self.last_received_segment = segments[-1]["text"]
+
+        if self.log_transcription:
+            # Truncate to last 3 entries for brevity.
+            text = text[-3:]
+            utils.clear_screen()
+            utils.print_transcript(text)
+
     def on_message(self, ws, message):
         """
         Callback function called when a message is received from the server.
-        
+
         It updates various attributes of the client based on the received message, including
         recording status, language detection, and server messages. If a disconnect message
         is received, it sets the recording status to False.
@@ -163,7 +148,6 @@ class Client:
             message (str): The received message from the server.
 
         """
-        self.last_response_recieved = time.time()
         message = json.loads(message)
 
         if self.uid != message.get("uid"):
@@ -171,21 +155,15 @@ class Client:
             return
 
         if "status" in message.keys():
-            if message["status"] == "WAIT":
-                self.waiting = True
-                print(
-                    f"[INFO]:Server is full. Estimated wait time {round(message['message'])} minutes."
-                )
-            elif message["status"] == "ERROR":
-                print(f"Message from Server: {message['message']}")
-                self.server_error = True
+            self.handle_status_messages(message)
             return
 
         if "message" in message.keys() and message["message"] == "DISCONNECT":
-            print("[INFO]: Server overtime disconnected.")
+            print("[INFO]: Server disconnected due to overtime.")
             self.recording = False
 
         if "message" in message.keys() and message["message"] == "SERVER_READY":
+            self.last_response_received = time.time()
             self.recording = True
             self.server_backend = message["backend"]
             print(f"[INFO]: Server Running with backend {self.server_backend}")
@@ -199,49 +177,23 @@ class Client:
             )
             return
 
-        if "segments" not in message.keys():
-            return
-
-        message = message["segments"]
-        text = []
-        n_segments = len(message)
-
-        if n_segments:
-            for i, seg in enumerate(message):
-                if text and text[-1] == seg["text"]:
-                    # already got it
-                    continue
-                text.append(seg["text"])
-
-                if i == n_segments-1: 
-                    self.last_segment = seg
-                elif self.server_backend == "faster_whisper":
-                    if not len(self.transcript) or float(seg['start']) >= float(self.transcript[-1]['end']):
-                        self.transcript.append(seg)
-
-        # keep only last 3
-        if len(text) > 3:
-            text = text[-3:]
-        wrapper = textwrap.TextWrapper(width=60)
-        word_list = wrapper.wrap(text="".join(text))
-        # Print each line.
-        if os.name == "nt":
-            os.system("cls")
-        else:
-            os.system("clear")
-        for element in word_list:
-            print(element)
+        if "segments" in message.keys():
+            self.process_segments(message["segments"])
 
     def on_error(self, ws, error):
-        print(error)
+        print(f"[ERROR] WebSocket Error: {error}")
+        self.server_error = True
+        self.error_message = error
 
     def on_close(self, ws, close_status_code, close_msg):
         print(f"[INFO]: Websocket connection closed: {close_status_code}: {close_msg}")
+        self.recording = False
+        self.waiting = False
 
     def on_open(self, ws):
         """
         Callback function called when the WebSocket connection is successfully opened.
-        
+
         Sends an initial configuration message to the server, including client UID,
         language selection, and task type.
 
@@ -257,27 +209,13 @@ class Client:
                     "language": self.language,
                     "task": self.task,
                     "model": self.model,
+                    "use_vad": self.use_vad,
+                    "max_clients": self.max_clients,
+                    "max_connection_time": self.max_connection_time,
                 }
             )
         )
 
-    @staticmethod
-    def bytes_to_float_array(audio_bytes):
-        """
-        Convert audio data from bytes to a NumPy float array.
-        
-        It assumes that the audio data is in 16-bit PCM format. The audio data is normalized to 
-        have values between -1 and 1.
-
-        Args:
-            audio_bytes (bytes): Audio data in bytes.
-
-        Returns:
-            np.ndarray: A NumPy array containing the audio data as float values normalized between -1 and 1.
-        """
-        raw_data = np.frombuffer(buffer=audio_bytes, dtype=np.int16)
-        return raw_data.astype(np.float32) / 32768.0
-
     def send_packet_to_server(self, message):
         """
         Send an audio packet to the server using WebSocket.
@@ -291,67 +229,11 @@ class Client:
         except Exception as e:
             print(e)
 
-    def play_file(self, filename):
-        """
-        Play an audio file and send it to the server for processing.
-        
-        Reads an audio file, plays it through the audio output, and simultaneously sends
-        the audio data to the server for processing. It uses PyAudio to create an audio 
-        stream for playback. The audio data is read from the file in chunks, converted to 
-        floating-point format, and sent to the server using WebSocket communication.
-        This method is typically used when you want to process pre-recorded audio and send it
-        to the server in real-time.
-
-        Args:
-            filename (str): The path to the audio file to be played and sent to the server.
-        """
-        
-        # read audio and create pyaudio stream
-        with wave.open(filename, "rb") as wavfile:
-            self.stream = self.p.open(
-                format=self.p.get_format_from_width(wavfile.getsampwidth()),
-                channels=wavfile.getnchannels(),
-                rate=wavfile.getframerate(),
-                input=True,
-                output=True,
-                frames_per_buffer=self.chunk,
-            )
-            try:
-                while self.recording:
-                    data = wavfile.readframes(self.chunk)
-                    if data == b"":
-                        break
-
-                    audio_array = self.bytes_to_float_array(data)
-                    self.send_packet_to_server(audio_array.tobytes())
-                    self.stream.write(data)
-
-                wavfile.close()
-
-                assert self.last_response_recieved
-                while time.time() - self.last_response_recieved < self.disconnect_if_no_response_for:
-                    continue
-
-                if self.server_backend == "faster_whisper":
-                    self.write_srt_file(self.srt_file_path)
-                self.stream.close()
-                self.close_websocket()
-
-            except KeyboardInterrupt:
-                wavfile.close()
-                self.stream.stop_stream()
-                self.stream.close()
-                self.p.terminate()
-                self.close_websocket()
-                if self.server_backend == "faster_whisper":
-                    self.write_srt_file(self.srt_file_path)
-                print("[INFO]: Keyboard interrupt.")
-
     def close_websocket(self):
         """
         Close the WebSocket connection and join the WebSocket thread.
 
-        First attempts to close the WebSocket connection using `self.client_socket.close()`. After 
+        First attempts to close the WebSocket connection using `self.client_socket.close()`. After
         closing the connection, it joins the WebSocket thread to ensure proper termination.
 
         """
@@ -374,11 +256,341 @@ class Client:
         """
         return self.client_socket
 
+    def write_srt_file(self, output_path="output.srt"):
+        """
+        Writes out the transcript in .srt format.
+
+        Args:
+            message (output_path, optional): The path to the target file.  Default is "output.srt".
+
+        """
+        if self.server_backend == "faster_whisper":
+            if not self.transcript and self.last_segment is not None:
+                self.transcript.append(self.last_segment)
+            elif self.last_segment and self.transcript[-1]["text"] != self.last_segment["text"]:
+                self.transcript.append(self.last_segment)
+            utils.create_srt_file(self.transcript, output_path)
+
+    def wait_before_disconnect(self):
+        """Waits a bit before disconnecting in order to process pending responses."""
+        assert self.last_response_received
+        while time.time() - self.last_response_received < self.disconnect_if_no_response_for:
+            continue
+
+
+class TranscriptionTeeClient:
+    """
+    Client for handling audio recording, streaming, and transcription tasks via one or more
+    WebSocket connections.
+
+    Acts as a high-level client for audio transcription tasks using a WebSocket connection. It can be used
+    to send audio data for transcription to one or more servers, and receive transcribed text segments.
+    Args:
+        clients (list): one or more previously initialized Client instances
+
+    Attributes:
+        clients (list): the underlying Client instances responsible for handling WebSocket connections.
+    """
+    def __init__(self, clients, save_output_recording=False, output_recording_filename="./output_recording.wav", mute_audio_playback=False):
+        self.clients = clients
+        if not self.clients:
+            raise Exception("At least one client is required.")
+        self.chunk = 4096
+        self.format = pyaudio.paInt16
+        self.channels = 1
+        self.rate = 16000
+        self.record_seconds = 60000
+        self.save_output_recording = save_output_recording
+        self.output_recording_filename = output_recording_filename
+        self.mute_audio_playback = mute_audio_playback
+        self.frames = b""
+        self.p = pyaudio.PyAudio()
+        try:
+            self.stream = self.p.open(
+                format=self.format,
+                channels=self.channels,
+                rate=self.rate,
+                input=True,
+                frames_per_buffer=self.chunk,
+            )
+        except OSError as error:
+            print(f"[WARN]: Unable to access microphone. {error}")
+            self.stream = None
+
+    def __call__(self, audio=None, rtsp_url=None, hls_url=None, save_file=None):
+        """
+        Start the transcription process.
+
+        Initiates the transcription process by connecting to the server via a WebSocket. It waits for the server
+        to be ready to receive audio data and then sends audio for transcription. If an audio file is provided, it
+        will be played and streamed to the server; otherwise, it will perform live recording.
+
+        Args:
+            audio (str, optional): Path to an audio file for transcription. Default is None, which triggers live recording.
+
+        """
+        assert sum(
+            source is not None for source in [audio, rtsp_url, hls_url]
+        ) <= 1, 'You must provide only one selected source'
+
+        print("[INFO]: Waiting for server ready ...")
+        for client in self.clients:
+            while not client.recording:
+                if client.waiting or client.server_error:
+                    self.close_all_clients()
+                    return
+
+        print("[INFO]: Server Ready!")
+        if hls_url is not None:
+            self.process_hls_stream(hls_url, save_file)
+        elif audio is not None:
+            resampled_file = utils.resample(audio)
+            self.play_file(resampled_file)
+        elif rtsp_url is not None:
+            self.process_rtsp_stream(rtsp_url)
+        else:
+            self.record()
+
+    def close_all_clients(self):
+        """Closes all client websockets."""
+        for client in self.clients:
+            client.close_websocket()
+
+    def write_all_clients_srt(self):
+        """Writes out .srt files for all clients."""
+        for client in self.clients:
+            client.write_srt_file(client.srt_file_path)
+
+    def multicast_packet(self, packet, unconditional=False):
+        """
+        Sends an identical packet via all clients.
+
+        Args:
+            packet (bytes): The audio data packet in bytes to be sent.
+            unconditional (bool, optional): If true, send regardless of whether clients are recording.  Default is False.
+        """
+        for client in self.clients:
+            if (unconditional or client.recording):
+                client.send_packet_to_server(packet)
+
+    def play_file(self, filename):
+        """
+        Play an audio file and send it to the server for processing.
+
+        Reads an audio file, plays it through the audio output, and simultaneously sends
+        the audio data to the server for processing. It uses PyAudio to create an audio
+        stream for playback. The audio data is read from the file in chunks, converted to
+        floating-point format, and sent to the server using WebSocket communication.
+        This method is typically used when you want to process pre-recorded audio and send it
+        to the server in real-time.
+
+        Args:
+            filename (str): The path to the audio file to be played and sent to the server.
+        """
+
+        # read audio and create pyaudio stream
+        with wave.open(filename, "rb") as wavfile:
+            self.stream = self.p.open(
+                format=self.p.get_format_from_width(wavfile.getsampwidth()),
+                channels=wavfile.getnchannels(),
+                rate=wavfile.getframerate(),
+                input=True,
+                output=True,
+                frames_per_buffer=self.chunk,
+            )
+            chunk_duration = self.chunk / float(wavfile.getframerate())
+            try:
+                while any(client.recording for client in self.clients):
+                    data = wavfile.readframes(self.chunk)
+                    if data == b"":
+                        break
+
+                    audio_array = self.bytes_to_float_array(data)
+                    self.multicast_packet(audio_array.tobytes())
+                    if self.mute_audio_playback:
+                        time.sleep(chunk_duration)
+                    else:
+                        self.stream.write(data)
+    
+                wavfile.close()
+
+                for client in self.clients:
+                    client.wait_before_disconnect()
+                self.multicast_packet(Client.END_OF_AUDIO.encode('utf-8'), True)
+                self.write_all_clients_srt()
+                self.stream.close()
+                self.close_all_clients()
+
+            except KeyboardInterrupt:
+                wavfile.close()
+                self.stream.stop_stream()
+                self.stream.close()
+                self.p.terminate()
+                self.close_all_clients()
+                self.write_all_clients_srt()
+                print("[INFO]: Keyboard interrupt.")
+
+    def process_rtsp_stream(self, rtsp_url):
+        """
+        Connect to an RTSP source, process the audio stream, and send it for transcription.
+
+        Args:
+            rtsp_url (str): The URL of the RTSP stream source.
+        """
+        print("[INFO]: Connecting to RTSP stream...")
+        try:
+            container = av.open(rtsp_url, format="rtsp", options={"rtsp_transport": "tcp"})
+            self.process_av_stream(container, stream_type="RTSP")
+        except Exception as e:
+            print(f"[ERROR]: Failed to process RTSP stream: {e}")
+        finally:
+            for client in self.clients:
+                client.wait_before_disconnect()
+            self.multicast_packet(Client.END_OF_AUDIO.encode('utf-8'), True)
+            self.close_all_clients()
+            self.write_all_clients_srt()
+        print("[INFO]: RTSP stream processing finished.")
+
+    def process_hls_stream(self, hls_url, save_file=None):
+        """
+        Connect to an HLS source, process the audio stream, and send it for transcription.
+
+        Args:
+            hls_url (str): The URL of the HLS stream source.
+            save_file (str, optional): Local path to save the network stream.
+        """
+        print("[INFO]: Connecting to HLS stream...")
+        try:
+            container = av.open(hls_url, format="hls")
+            self.process_av_stream(container, stream_type="HLS", save_file=save_file)
+        except Exception as e:
+            print(f"[ERROR]: Failed to process HLS stream: {e}")
+        finally:
+            for client in self.clients:
+                client.wait_before_disconnect()
+            self.multicast_packet(Client.END_OF_AUDIO.encode('utf-8'), True)
+            self.close_all_clients()
+            self.write_all_clients_srt()
+        print("[INFO]: HLS stream processing finished.")
+
+    def process_av_stream(self, container, stream_type, save_file=None):
+        """
+        Process an AV container stream and send audio packets to the server.
+
+        Args:
+            container (av.container.InputContainer): The input container to process.
+            stream_type (str): The type of stream being processed ("RTSP" or "HLS").
+            save_file (str, optional): Local path to save the stream. Default is None.
+        """
+        audio_stream = next((s for s in container.streams if s.type == "audio"), None)
+        if not audio_stream:
+            print(f"[ERROR]: No audio stream found in {stream_type} source.")
+            return
+
+        output_container = None
+        if save_file:
+            output_container = av.open(save_file, mode="w")
+            output_audio_stream = output_container.add_stream(codec_name="pcm_s16le", rate=self.rate)
+
+        try:
+            for packet in container.demux(audio_stream):
+                for frame in packet.decode():
+                    audio_data = frame.to_ndarray().tobytes()
+                    self.multicast_packet(audio_data)
+
+                    if save_file:
+                        output_container.mux(frame)
+        except Exception as e:
+            print(f"[ERROR]: Error during {stream_type} stream processing: {e}")
+        finally:
+            # Wait for server to send any leftover transcription.
+            time.sleep(5)
+            self.multicast_packet(Client.END_OF_AUDIO.encode('utf-8'), True)
+            if output_container:
+                output_container.close()
+            container.close()
+
+    def save_chunk(self, n_audio_file):
+        """
+        Saves the current audio frames to a WAV file in a separate thread.
+
+        Args:
+        n_audio_file (int): The index of the audio file which determines the filename.
+                            This helps in maintaining the order and uniqueness of each chunk.
+        """
+        t = threading.Thread(
+            target=self.write_audio_frames_to_file,
+            args=(self.frames[:], f"chunks/{n_audio_file}.wav",),
+        )
+        t.start()
+
+    def finalize_recording(self, n_audio_file):
+        """
+        Finalizes the recording process by saving any remaining audio frames,
+        closing the audio stream, and terminating the process.
+
+        Args:
+        n_audio_file (int): The file index to be used if there are remaining audio frames to be saved.
+                            This index is incremented before use if the last chunk is saved.
+        """
+        if self.save_output_recording and len(self.frames):
+            self.write_audio_frames_to_file(
+                self.frames[:], f"chunks/{n_audio_file}.wav"
+            )
+            n_audio_file += 1
+        self.stream.stop_stream()
+        self.stream.close()
+        self.p.terminate()
+        self.close_all_clients()
+        if self.save_output_recording:
+            self.write_output_recording(n_audio_file)
+        self.write_all_clients_srt()
+
+    def record(self):
+        """
+        Record audio data from the input stream and save it to a WAV file.
+
+        Continuously records audio data from the input stream, sends it to the server via a WebSocket
+        connection, and simultaneously saves it to multiple WAV files in chunks. It stops recording when
+        the `RECORD_SECONDS` duration is reached or when the `RECORDING` flag is set to `False`.
+
+        Audio data is saved in chunks to the "chunks" directory. Each chunk is saved as a separate WAV file.
+        The recording will continue until the specified duration is reached or until the `RECORDING` flag is set to `False`.
+        The recording process can be interrupted by sending a KeyboardInterrupt (e.g., pressing Ctrl+C). After recording,
+        the method combines all the saved audio chunks into the specified `out_file`.
+        """
+        n_audio_file = 0
+        if self.save_output_recording:
+            if os.path.exists("chunks"):
+                shutil.rmtree("chunks")
+            os.makedirs("chunks")
+        try:
+            for _ in range(0, int(self.rate / self.chunk * self.record_seconds)):
+                if not any(client.recording for client in self.clients):
+                    break
+                data = self.stream.read(self.chunk, exception_on_overflow=False)
+                self.frames += data
+
+                audio_array = self.bytes_to_float_array(data)
+
+                self.multicast_packet(audio_array.tobytes())
+
+                # save frames if more than a minute
+                if len(self.frames) > 60 * self.rate:
+                    if self.save_output_recording:
+                        self.save_chunk(n_audio_file)
+                        n_audio_file += 1
+                    self.frames = b""
+            self.write_all_clients_srt()
+
+        except KeyboardInterrupt:
+            self.finalize_recording(n_audio_file)
+
     def write_audio_frames_to_file(self, frames, file_name):
         """
         Write audio frames to a WAV file.
 
-        The WAV file is created or overwritten with the specified name. The audio frames should be 
+        The WAV file is created or overwritten with the specified name. The audio frames should be
         in the correct format and match the specified channel, sample width, and sample rate.
 
         Args:
@@ -393,108 +605,11 @@ class Client:
             wavfile.setframerate(self.rate)
             wavfile.writeframes(frames)
 
-    def process_hls_stream(self, hls_url):
-        """
-        Connect to an HLS source, process the audio stream, and send it for transcription.
-
-        Args:
-            hls_url (str): The URL of the HLS stream source.
-        """
-        print("[INFO]: Connecting to HLS stream...")
-        process = None  # Initialize process to None
-
-        try:
-            # Connecting to the HLS stream using ffmpeg-python
-            process = (
-                ffmpeg
-                .input(hls_url, threads=0)
-                .output('-', format='s16le', acodec='pcm_s16le', ac=1, ar=self.rate)
-                .run_async(pipe_stdout=True, pipe_stderr=True)
-            )
-
-            # Process the stream
-            while True:
-                in_bytes = process.stdout.read(self.chunk * 2)  # 2 bytes per sample
-                if not in_bytes:
-                    break
-                audio_array = self.bytes_to_float_array(in_bytes)
-                self.send_packet_to_server(audio_array.tobytes())
-
-        except Exception as e:
-            print(f"[ERROR]: Failed to connect to HLS stream: {e}")
-        finally:
-            if process:
-                process.kill()
-
-        print("[INFO]: HLS stream processing finished.")
-
-
-    def record(self, out_file="output_recording.wav"):
-        """
-        Record audio data from the input stream and save it to a WAV file.
-
-        Continuously records audio data from the input stream, sends it to the server via a WebSocket
-        connection, and simultaneously saves it to multiple WAV files in chunks. It stops recording when
-        the `RECORD_SECONDS` duration is reached or when the `RECORDING` flag is set to `False`.
-
-        Audio data is saved in chunks to the "chunks" directory. Each chunk is saved as a separate WAV file.
-        The recording will continue until the specified duration is reached or until the `RECORDING` flag is set to `False`.
-        The recording process can be interrupted by sending a KeyboardInterrupt (e.g., pressing Ctrl+C). After recording, 
-        the method combines all the saved audio chunks into the specified `out_file`.
-
-        Args:
-            out_file (str, optional): The name of the output WAV file to save the entire recording. Default is "output_recording.wav".
-
-        """
-        n_audio_file = 0
-        if not os.path.exists("chunks"):
-            os.makedirs("chunks", exist_ok=True)
-        try:
-            for _ in range(0, int(self.rate / self.chunk * self.record_seconds)):
-                if not self.recording:
-                    break
-                data = self.stream.read(self.chunk, exception_on_overflow = False)
-                self.frames += data
-
-                audio_array = Client.bytes_to_float_array(data)
-
-                self.send_packet_to_server(audio_array.tobytes())
-
-                # save frames if more than a minute
-                if len(self.frames) > 60 * self.rate:
-                    t = threading.Thread(
-                        target=self.write_audio_frames_to_file,
-                        args=(
-                            self.frames[:],
-                            f"chunks/{n_audio_file}.wav",
-                        ),
-                    )
-                    t.start()
-                    n_audio_file += 1
-                    self.frames = b""
-            if self.server_backend == "faster_whisper":
-                self.write_srt_file(self.srt_file_path)
-
-        except KeyboardInterrupt:
-            if len(self.frames):
-                self.write_audio_frames_to_file(
-                    self.frames[:], f"chunks/{n_audio_file}.wav"
-                )
-                n_audio_file += 1
-            self.stream.stop_stream()
-            self.stream.close()
-            self.p.terminate()
-            self.close_websocket()
-
-            self.write_output_recording(n_audio_file, out_file)
-            if self.server_backend == "faster_whisper":
-                self.write_srt_file(self.srt_file_path)
-
-    def write_output_recording(self, n_audio_file, out_file):
+    def write_output_recording(self, n_audio_file):
         """
         Combine and save recorded audio chunks into a single WAV file.
-        
-        The individual audio chunk files are expected to be located in the "chunks" directory. Reads each chunk 
+
+        The individual audio chunk files are expected to be located in the "chunks" directory. Reads each chunk
         file, appends its audio data to the final recording, and then deletes the chunk file. After combining
         and saving, the final recording is stored in the specified `out_file`.
 
@@ -509,7 +624,7 @@ class Client:
             for i in range(n_audio_file)
             if os.path.exists(f"chunks/{i}.wav")
         ]
-        with wave.open(out_file, "wb") as wavfile:
+        with wave.open(self.output_recording_filename, "wb") as wavfile:
             wavfile: wave.Wave_write
             wavfile.setnchannels(self.channels)
             wavfile.setsampwidth(2)
@@ -524,15 +639,31 @@ class Client:
                 # remove this file
                 os.remove(in_file)
         wavfile.close()
+        # clean up temporary directory to store chunks
+        if os.path.exists("chunks"):
+            shutil.rmtree("chunks")
 
-    def write_srt_file(self, output_path="output.srt"):
-        self.transcript.append(self.last_segment)
-        create_srt_file(self.transcript, output_path)
+    @staticmethod
+    def bytes_to_float_array(audio_bytes):
+        """
+        Convert audio data from bytes to a NumPy float array.
+
+        It assumes that the audio data is in 16-bit PCM format. The audio data is normalized to
+        have values between -1 and 1.
+
+        Args:
+            audio_bytes (bytes): Audio data in bytes.
+
+        Returns:
+            np.ndarray: A NumPy array containing the audio data as float values normalized between -1 and 1.
+        """
+        raw_data = np.frombuffer(buffer=audio_bytes, dtype=np.int16)
+        return raw_data.astype(np.float32) / 32768.0
 
 
-class TranscriptionClient:
+class TranscriptionClient(TranscriptionTeeClient):
     """
-    Client for handling audio transcription tasks via a WebSocket connection.
+    Client for handling audio transcription tasks via a single WebSocket connection.
 
     Acts as a high-level client for audio transcription tasks using a WebSocket connection. It can be used
     to send audio data for transcription to a server and receive transcribed text segments.
@@ -541,7 +672,16 @@ class TranscriptionClient:
         host (str): The hostname or IP address of the server.
         port (int): The port number to connect to on the server.
         lang (str, optional): The primary language for transcription. Default is None, which defaults to English ('en').
-        translate (bool, optional): Indicates whether translation tasks are required (default is False).
+        translate (bool, optional): If True, the task will be translation instead of transcription. Default is False.
+        model (str, optional): The whisper model to use (e.g., "small", "base"). Default is "small".
+        use_vad (bool, optional): Whether to enable voice activity detection. Default is True.
+        save_output_recording (bool, optional): Whether to save the microphone recording. Default is False.
+        output_recording_filename (str, optional): Path to save the output recording WAV file. Default is "./output_recording.wav".
+        output_transcription_path (str, optional): File path to save the output transcription (SRT file). Default is "./output.srt".
+        log_transcription (bool, optional): Whether to log transcription output to the console. Default is True.
+        max_clients (int, optional): Maximum number of client connections allowed. Default is 4.
+        max_connection_time (int, optional): Maximum allowed connection time in seconds. Default is 600.
+        mute_audio_playback (bool, optional): If True, mutes audio playback during file playback. Default is False.
 
     Attributes:
         client (Client): An instance of the underlying Client class responsible for handling the WebSocket connection.
@@ -553,38 +693,36 @@ class TranscriptionClient:
         transcription_client()
         ```
     """
-    def __init__(self,
+    def __init__(
+        self,
         host,
         port,
         lang=None,
         translate=False,
         model="small",
+        use_vad=True,
+        save_output_recording=False,
+        output_recording_filename="./output_recording.wav",
+        output_transcription_path="./output.srt",
+        log_transcription=True,
+        max_clients=4,
+        max_connection_time=600,
+        mute_audio_playback=False,
     ):
-        self.client = Client(host, port, lang, translate, model)
+        self.client = Client(
+            host, port, lang, translate, model, srt_file_path=output_transcription_path,
+            use_vad=use_vad, log_transcription=log_transcription, max_clients=max_clients,
+            max_connection_time=max_connection_time
+        )
 
-    def __call__(self, audio=None, hls_url=None):
-        """
-        Start the transcription process.
-
-        Initiates the transcription process by connecting to the server via a WebSocket. It waits for the server
-        to be ready to receive audio data and then sends audio for transcription. If an audio file is provided, it 
-        will be played and streamed to the server; otherwise, it will perform live recording.
-
-        Args:
-            audio (str, optional): Path to an audio file for transcription. Default is None, which triggers live recording.
-                   
-        """
-        print("[INFO]: Waiting for server ready ...")
-        while not self.client.recording:
-            if self.client.waiting or self.client.server_error:
-                self.client.close_websocket()
-                return
-
-        print("[INFO]: Server Ready!")
-        if hls_url is not None:
-            self.client.process_hls_stream(hls_url)
-        elif audio is not None:
-            resampled_file = resample(audio)
-            self.client.play_file(resampled_file)
-        else:
-            self.client.record()
+        if save_output_recording and not output_recording_filename.endswith(".wav"):
+            raise ValueError(f"Please provide a valid `output_recording_filename`: {output_recording_filename}")
+        if not output_transcription_path.endswith(".srt"):
+            raise ValueError(f"Please provide a valid `output_transcription_path`: {output_transcription_path}. The file extension should be `.srt`.")
+        TranscriptionTeeClient.__init__(
+            self,
+            [self.client],
+            save_output_recording=save_output_recording,
+            output_recording_filename=output_recording_filename,
+            mute_audio_playback=mute_audio_playback
+        )

whisper_live/tensorrt_utils.py

@@ -23,8 +23,12 @@ from typing import Dict, Iterable, List, Optional, TextIO, Tuple, Union
 import kaldialign
 import numpy as np
 import soundfile
+import av
+import wave
 import torch
 import torch.nn.functional as F
+from whisper_live.utils import resample
+
 
 Pathlike = Union[str, Path]
 
@@ -35,38 +39,33 @@ CHUNK_LENGTH = 30
 N_SAMPLES = CHUNK_LENGTH * SAMPLE_RATE  # 480000 samples in a 30-second chunk
 
 
-def load_audio(file: str, sr: int = SAMPLE_RATE):
+def load_audio(file: str, sr: int = 16000):
     """
-    Open an audio file and read as mono waveform, resampling as necessary
+    Open an audio file, resample it, and read as a mono waveform.
 
     Parameters
     ----------
     file: str
-        The audio file to open
+        The audio file to open.
 
     sr: int
-        The sample rate to resample the audio if necessary
+        The sample rate to resample the audio if necessary.
 
     Returns
     -------
     A NumPy array containing the audio waveform, in float32 dtype.
     """
+    resampled_file = resample(file, sr)
 
-    # This launches a subprocess to decode audio while down-mixing
-    # and resampling as necessary.  Requires the ffmpeg CLI in PATH.
-    # fmt: off
-    cmd = [
-        "ffmpeg", "-nostdin", "-threads", "0", "-i", file, "-f", "s16le", "-ac",
-        "1", "-acodec", "pcm_s16le", "-ar",
-        str(sr), "-"
-    ]
-    # fmt: on
-    try:
-        out = run(cmd, capture_output=True, check=True).stdout
-    except CalledProcessError as e:
-        raise RuntimeError(f"Failed to load audio: {e.stderr.decode()}") from e
+    with wave.open(resampled_file, "rb") as wav_file:
+        num_frames = wav_file.getnframes()
+        raw_data = wav_file.readframes(num_frames)
 
-    return np.frombuffer(out, np.int16).flatten().astype(np.float32) / 32768.0
+        audio_data = np.frombuffer(raw_data, dtype=np.int16)
+
+    audio_data = audio_data.astype(np.float32) / 32768.0
+
+    return audio_data
 
 
 def load_audio_wav_format(wav_path):
@@ -214,7 +213,7 @@ def store_transcripts(filename: Pathlike, texts: Iterable[Tuple[str, str,
             print(f"{cut_id}:\thyp={hyp}", file=f)
 
 
-def write_error_stats(
+def write_error_stats(                                              # noqa: C901
     f: TextIO,
     test_set_name: str,
     results: List[Tuple[str, str]],
@@ -362,4 +361,4 @@ def write_error_stats(
         hyp_count = corr + hyp_sub + ins
 
         print(f"{word}   {corr} {tot_errs} {ref_count} {hyp_count}", file=f)
-    return float(tot_err_rate)
+    return float(tot_err_rate)

whisper_live/transcriber_tensorrt.py

@@ -1,23 +1,22 @@
-import argparse
 import json
 import re
-import time
+import math
 from collections import OrderedDict
 from pathlib import Path
-from typing import Dict, Iterable, List, Optional, TextIO, Tuple, Union
+from typing import Union
 
 import torch
 import numpy as np
+import torch.nn.functional as F
 from whisper.tokenizer import get_tokenizer
-from whisper_live.tensorrt_utils import (mel_filters, store_transcripts,
-                           write_error_stats, load_audio_wav_format,
-                           pad_or_trim, load_audio)
+from whisper_live.tensorrt_utils import (mel_filters, load_audio_wav_format, pad_or_trim, load_audio)
 
 import tensorrt_llm
 import tensorrt_llm.logger as logger
 from tensorrt_llm._utils import (str_dtype_to_torch, str_dtype_to_trt,
                                  trt_dtype_to_torch)
-from tensorrt_llm.runtime import ModelConfig, SamplingConfig
+from tensorrt_llm.bindings import GptJsonConfig, KVCacheType
+from tensorrt_llm.runtime import PYTHON_BINDINGS, ModelConfig, SamplingConfig
 from tensorrt_llm.runtime.session import Session, TensorInfo
 
 
@@ -27,41 +26,102 @@ HOP_LENGTH = 160
 CHUNK_LENGTH = 30
 N_SAMPLES = CHUNK_LENGTH * SAMPLE_RATE  # 480000 samples in a 30-second chunk
 
+def read_config(component, engine_dir):
+    config_path = engine_dir / component / 'config.json'
+    with open(config_path, 'r') as f:
+        config = json.load(f)
+    model_config = OrderedDict()
+    model_config.update(config['pretrained_config'])
+    model_config.update(config['build_config'])
+    return model_config
+
+
+def remove_tensor_padding(input_tensor,
+                          input_tensor_lengths=None,
+                          pad_value=None):
+    if pad_value:
+        assert input_tensor_lengths is None, "input_tensor_lengths should be None when pad_value is provided"
+        # Text tensor case: batch, seq_len
+        assert torch.all(
+            input_tensor[:, 0] != pad_value
+        ), "First token in each sequence should not be pad_value"
+        assert input_tensor_lengths is None
+
+        # Create a mask for all non-pad tokens
+        mask = input_tensor != pad_value
+
+        # Apply the mask to input_tensor to remove pad tokens
+        output_tensor = input_tensor[mask].view(1, -1)
+
+    else:
+        # Audio tensor case: batch, seq_len, feature_len
+        # position_ids case: batch, seq_len
+        assert input_tensor_lengths is not None, "input_tensor_lengths must be provided for 3D input_tensor"
+
+        # Initialize a list to collect valid sequences
+        valid_sequences = []
+
+        for i in range(input_tensor.shape[0]):
+            valid_length = input_tensor_lengths[i]
+            valid_sequences.append(input_tensor[i, :valid_length])
+
+        # Concatenate all valid sequences along the batch dimension
+        output_tensor = torch.cat(valid_sequences, dim=0)
+    return output_tensor
+
 
 class WhisperEncoding:
 
     def __init__(self, engine_dir):
         self.session = self.get_session(engine_dir)
+        config = read_config('encoder', engine_dir)
+        self.n_mels = config['n_mels']
+        self.dtype = config['dtype']
+        self.num_languages = config['num_languages']
+        self.encoder_config = config
 
     def get_session(self, engine_dir):
-        config_path = engine_dir / 'encoder_config.json'
-        with open(config_path, 'r') as f:
-            config = json.load(f)
-
-        use_gpt_attention_plugin = config['plugin_config'][
-            'gpt_attention_plugin']
-        dtype = config['builder_config']['precision']
-        n_mels = config['builder_config']['n_mels']
-        num_languages = config['builder_config']['num_languages']
-
-        self.dtype = dtype
-        self.n_mels = n_mels
-        self.num_languages = num_languages
-
-        serialize_path = engine_dir / f'whisper_encoder_{self.dtype}_tp1_rank0.engine'
-
+        serialize_path = engine_dir / 'encoder' / 'rank0.engine'
         with open(serialize_path, 'rb') as f:
             session = Session.from_serialized_engine(f.read())
-
         return session
 
-    def get_audio_features(self, mel):
+    def get_audio_features(self,
+                           mel,
+                           mel_input_lengths,
+                           encoder_downsampling_factor=2):
+        if isinstance(mel, list):
+            longest_mel = max([f.shape[-1] for f in mel])
+            mel = [
+                torch.nn.functional.pad(f, (0, longest_mel - f.shape[-1]),
+                                        mode='constant') for f in mel
+            ]
+            mel = torch.cat(mel, dim=0).type(
+                str_dtype_to_torch("float16")).contiguous()
+        bsz, seq_len = mel.shape[0], mel.shape[2]
+        position_ids = torch.arange(
+            math.ceil(seq_len / encoder_downsampling_factor),
+            dtype=torch.int32,
+            device=mel.device).expand(bsz, -1).contiguous()
+        if self.encoder_config['plugin_config']['remove_input_padding']:
+            # mel B,D,T -> B,T,D -> BxT, D
+            mel = mel.transpose(1, 2)
+            mel = remove_tensor_padding(mel, mel_input_lengths)
+            position_ids = remove_tensor_padding(
+                position_ids, mel_input_lengths // encoder_downsampling_factor)
         inputs = OrderedDict()
-        output_list = []
+        inputs['input_features'] = mel
+        inputs['input_lengths'] = mel_input_lengths
+        inputs['position_ids'] = position_ids
 
-        inputs.update({'x': mel})
-        output_list.append(
-            TensorInfo('x', str_dtype_to_trt(self.dtype), mel.shape))
+        output_list = [
+            TensorInfo('input_features', str_dtype_to_trt(self.dtype),
+                       mel.shape),
+            TensorInfo('input_lengths', str_dtype_to_trt('int32'),
+                       mel_input_lengths.shape),
+            TensorInfo('position_ids', str_dtype_to_trt('int32'),
+                       inputs['position_ids'].shape)
+        ]
 
         output_info = (self.session).infer_shapes(output_list)
 
@@ -78,46 +138,44 @@ class WhisperEncoding:
                               stream=stream.cuda_stream)
         assert ok, 'Engine execution failed'
         stream.synchronize()
-        audio_features = outputs['output']
-        return audio_features
+        encoder_output = outputs['encoder_output']
+        encoder_output_lengths = mel_input_lengths // encoder_downsampling_factor
+        return encoder_output, encoder_output_lengths
 
 
 class WhisperDecoding:
 
     def __init__(self, engine_dir, runtime_mapping, debug_mode=False):
 
-        self.decoder_config = self.get_config(engine_dir)
+        self.decoder_config = read_config('decoder', engine_dir)
         self.decoder_generation_session = self.get_session(
             engine_dir, runtime_mapping, debug_mode)
 
-    def get_config(self, engine_dir):
-        config_path = engine_dir / 'decoder_config.json'
-        with open(config_path, 'r') as f:
-            config = json.load(f)
-        decoder_config = OrderedDict()
-        decoder_config.update(config['plugin_config'])
-        decoder_config.update(config['builder_config'])
-        return decoder_config
-
     def get_session(self, engine_dir, runtime_mapping, debug_mode=False):
-        dtype = self.decoder_config['precision']
-        serialize_path = engine_dir / f'whisper_decoder_{dtype}_tp1_rank0.engine'
+        serialize_path = engine_dir / 'decoder' / 'rank0.engine'
         with open(serialize_path, "rb") as f:
             decoder_engine_buffer = f.read()
 
         decoder_model_config = ModelConfig(
-            num_heads=self.decoder_config['num_heads'],
-            num_kv_heads=self.decoder_config['num_heads'],
+            max_batch_size=self.decoder_config['max_batch_size'],
+            max_beam_width=self.decoder_config['max_beam_width'],
+            num_heads=self.decoder_config['num_attention_heads'],
+            num_kv_heads=self.decoder_config['num_attention_heads'],
             hidden_size=self.decoder_config['hidden_size'],
             vocab_size=self.decoder_config['vocab_size'],
-            num_layers=self.decoder_config['num_layers'],
-            gpt_attention_plugin=self.decoder_config['gpt_attention_plugin'],
-            remove_input_padding=self.decoder_config['remove_input_padding'],
-            cross_attention=self.decoder_config['cross_attention'],
+            cross_attention=True,
+            num_layers=self.decoder_config['num_hidden_layers'],
+            gpt_attention_plugin=self.decoder_config['plugin_config']
+            ['gpt_attention_plugin'],
+            remove_input_padding=self.decoder_config['plugin_config']
+            ['remove_input_padding'],
+            kv_cache_type=KVCacheType.PAGED
+            if self.decoder_config['plugin_config']['paged_kv_cache'] == True
+            else KVCacheType.CONTINUOUS,
             has_position_embedding=self.
             decoder_config['has_position_embedding'],
-            has_token_type_embedding=self.
-            decoder_config['has_token_type_embedding'],
+            dtype=self.decoder_config['dtype'],
+            has_token_type_embedding=False,
         )
         decoder_generation_session = tensorrt_llm.runtime.GenerationSession(
             decoder_model_config,
@@ -130,14 +188,12 @@ class WhisperDecoding:
     def generate(self,
                  decoder_input_ids,
                  encoder_outputs,
+                 encoder_max_input_length,
+                 encoder_input_lengths,
                  eot_id,
                  max_new_tokens=40,
                  num_beams=1):
-        encoder_input_lengths = torch.tensor(
-            [encoder_outputs.shape[1] for x in range(encoder_outputs.shape[0])],
-            dtype=torch.int32,
-            device='cuda')
-
+        batch_size = decoder_input_ids.shape[0]
         decoder_input_lengths = torch.tensor([
             decoder_input_ids.shape[-1]
             for _ in range(decoder_input_ids.shape[0])
@@ -146,6 +202,10 @@ class WhisperDecoding:
                                              device='cuda')
         decoder_max_input_length = torch.max(decoder_input_lengths).item()
 
+        cross_attention_mask = torch.ones([
+            batch_size, decoder_max_input_length + max_new_tokens,
+            encoder_max_input_length
+        ]).int().cuda()
         # generation config
         sampling_config = SamplingConfig(end_id=eot_id,
                                          pad_id=eot_id,
@@ -155,17 +215,31 @@ class WhisperDecoding:
             decoder_max_input_length,
             max_new_tokens,
             beam_width=num_beams,
-            encoder_max_input_length=encoder_outputs.shape[1])
+            encoder_max_input_length=encoder_max_input_length)
 
         torch.cuda.synchronize()
 
         decoder_input_ids = decoder_input_ids.type(torch.int32).cuda()
+        if self.decoder_config['plugin_config']['remove_input_padding']:
+            # 50256 is the index of <pad> for all whisper models' decoder
+            WHISPER_PAD_TOKEN_ID = 50256
+            decoder_input_ids = remove_tensor_padding(
+                decoder_input_ids, pad_value=WHISPER_PAD_TOKEN_ID)
+            if encoder_outputs.dim() == 3:
+                encoder_output_lens = torch.full((encoder_outputs.shape[0], ),
+                                                 encoder_outputs.shape[1],
+                                                 dtype=torch.int32,
+                                                 device='cuda')
+
+                encoder_outputs = remove_tensor_padding(encoder_outputs,
+                                                        encoder_output_lens)
         output_ids = self.decoder_generation_session.decode(
             decoder_input_ids,
             decoder_input_lengths,
             sampling_config,
             encoder_output=encoder_outputs,
             encoder_input_lengths=encoder_input_lengths,
+            cross_attention_mask=cross_attention_mask,
         )
         torch.cuda.synchronize()
 
@@ -176,33 +250,30 @@ class WhisperDecoding:
 
 class WhisperTRTLLM(object):
 
-    def __init__(
-        self,
-        engine_dir,
-        debug_mode=False,
-        assets_dir=None,
-        device=None,
-        is_multilingual=False,
-        language="en",
-        task="transcribe"
-        ):
+    def __init__(self, engine_dir, assets_dir=None, device=None, is_multilingual=False,
+                 language="en", task="transcribe"):
         world_size = 1
         runtime_rank = tensorrt_llm.mpi_rank()
         runtime_mapping = tensorrt_llm.Mapping(world_size, runtime_rank)
         torch.cuda.set_device(runtime_rank % runtime_mapping.gpus_per_node)
         engine_dir = Path(engine_dir)
+        encoder_config = read_config('encoder', engine_dir)
+        decoder_config = read_config('decoder', engine_dir)
+        self.n_mels = encoder_config['n_mels']
+        self.num_languages = encoder_config['num_languages']
+        is_multilingual = (decoder_config['vocab_size'] >= 51865)
 
         self.encoder = WhisperEncoding(engine_dir)
         self.decoder = WhisperDecoding(engine_dir,
-                                       runtime_mapping,
-                                       debug_mode=False)
+                                        runtime_mapping,
+                                        debug_mode=False)
         self.n_mels = self.encoder.n_mels
         # self.tokenizer = get_tokenizer(num_languages=self.encoder.num_languages,
         #                                tokenizer_dir=assets_dir)
         self.device = device
         self.tokenizer = get_tokenizer(
             is_multilingual,
-            num_languages=self.encoder.num_languages,
+            num_languages=self.num_languages,
             language=language,
             task=task,
         )
@@ -212,7 +283,7 @@ class WhisperTRTLLM(object):
         self,
         audio: Union[str, np.ndarray, torch.Tensor],
         padding: int = 0,
-        return_duration = True
+        return_duration=True
     ):
         """
         Compute the log-Mel spectrogram of
@@ -242,8 +313,7 @@ class WhisperTRTLLM(object):
                     audio, _ = load_audio_wav_format(audio)
                 else:
                     audio = load_audio(audio)
-            assert isinstance(audio,
-                            np.ndarray), f"Unsupported audio type: {type(audio)}"
+            assert isinstance(audio, np.ndarray), f"Unsupported audio type: {type(audio)}"
             duration = audio.shape[-1] / SAMPLE_RATE
             audio = pad_or_trim(audio, N_SAMPLES)
             audio = audio.astype(np.float32)
@@ -254,14 +324,9 @@ class WhisperTRTLLM(object):
         if padding > 0:
             audio = F.pad(audio, (0, padding))
         window = torch.hann_window(N_FFT).to(audio.device)
-        stft = torch.stft(audio,
-                        N_FFT,
-                        HOP_LENGTH,
-                        window=window,
-                        return_complex=True)
+        stft = torch.stft(audio, N_FFT, HOP_LENGTH, window=window, return_complex=True)
         magnitudes = stft[..., :-1].abs()**2
 
-        
         mel_spec = self.filters @ magnitudes
 
         log_spec = torch.clamp(mel_spec, min=1e-10).log10()
@@ -272,12 +337,13 @@ class WhisperTRTLLM(object):
         else:
             return log_spec
 
-
     def process_batch(
             self,
             mel,
+            mel_input_lengths,
             text_prefix="<|startoftranscript|><|en|><|transcribe|><|notimestamps|>",
-            num_beams=1):
+            num_beams=1,
+            max_new_tokens=96):
         prompt_id = self.tokenizer.encode(
             text_prefix, allowed_special=set(self.tokenizer.special_tokens.keys()))
 
@@ -285,18 +351,21 @@ class WhisperTRTLLM(object):
         batch_size = mel.shape[0]
         decoder_input_ids = prompt_id.repeat(batch_size, 1)
 
-        encoder_output = self.encoder.get_audio_features(mel)
+        encoder_output, encoder_output_lengths = self.encoder.get_audio_features(mel, mel_input_lengths)
+        encoder_max_input_length = torch.max(encoder_output_lengths).item()
         output_ids = self.decoder.generate(decoder_input_ids,
                                            encoder_output,
+                                           encoder_max_input_length,
+                                           encoder_output_lengths,
                                            self.tokenizer.eot,
-                                           max_new_tokens=96,
+                                           max_new_tokens=max_new_tokens,
                                            num_beams=num_beams)
         texts = []
         for i in range(len(output_ids)):
             text = self.tokenizer.decode(output_ids[i][0]).strip()
             texts.append(text)
         return texts
-    
+
     def transcribe(
             self,
             mel,
@@ -304,10 +373,22 @@ class WhisperTRTLLM(object):
             dtype='float16',
             batch_size=1,
             num_beams=1,
+            padding_strategy="max",
             ):
         mel = mel.type(str_dtype_to_torch(dtype))
         mel = mel.unsqueeze(0)
-        predictions = self.process_batch(mel, text_prefix, num_beams)
+        # repeat the mel spectrogram to match the batch size
+        mel = mel.repeat(batch_size, 1, 1)
+        if padding_strategy == "longest":
+            pass
+        else:
+            mel = torch.nn.functional.pad(mel, (0, 3000 - mel.shape[2]))
+        features_input_lengths = torch.full((mel.shape[0], ),
+                                             mel.shape[2],
+                                             dtype=torch.int32,
+                                             device=mel.device)
+
+        predictions = self.process_batch(mel, features_input_lengths, text_prefix, num_beams)
         prediction = predictions[0]
 
         # remove all special tokens in the prediction
@@ -336,5 +417,5 @@ def decode_wav_file(
     prediction = re.sub(r'<\|.*?\|>', '', prediction)
     if normalizer:
         prediction = normalizer(prediction)
-    
+
     return prediction.strip()

whisper_live/utils.py

@@ -0,0 +1,82 @@
+import os
+import textwrap
+import scipy
+import numpy as np
+import av
+from pathlib import Path
+
+
+def clear_screen():
+    """Clears the console screen."""
+    os.system("cls" if os.name == "nt" else "clear")
+
+
+def print_transcript(text):
+    """Prints formatted transcript text."""
+    wrapper = textwrap.TextWrapper(width=60)
+    for line in wrapper.wrap(text="".join(text)):
+        print(line)
+
+
+def format_time(s):
+    """Convert seconds (float) to SRT time format."""
+    hours = int(s // 3600)
+    minutes = int((s % 3600) // 60)
+    seconds = int(s % 60)
+    milliseconds = int((s - int(s)) * 1000)
+    return f"{hours:02}:{minutes:02}:{seconds:02},{milliseconds:03}"
+
+
+def create_srt_file(segments, resampled_file):
+    with open(resampled_file, 'w', encoding='utf-8') as srt_file:
+        segment_number = 1
+        for segment in segments:
+            start_time = format_time(float(segment['start']))
+            end_time = format_time(float(segment['end']))
+            text = segment['text']
+
+            srt_file.write(f"{segment_number}\n")
+            srt_file.write(f"{start_time} --> {end_time}\n")
+            srt_file.write(f"{text}\n\n")
+
+            segment_number += 1
+
+
+def resample(file: str, sr: int = 16000):
+    """
+    Resample the audio file to 16kHz.
+
+    Args:
+        file (str): The audio file to open
+        sr (int): The sample rate to resample the audio if necessary
+
+    Returns:
+        resampled_file (str): The resampled audio file
+    """
+    container = av.open(file)
+    stream = next(s for s in container.streams if s.type == 'audio')
+
+    resampler = av.AudioResampler(
+        format='s16',
+        layout='mono',
+        rate=sr,
+    )
+
+    resampled_file = Path(file).stem + "_resampled.wav"
+    output_container = av.open(resampled_file, mode='w')
+    output_stream = output_container.add_stream('pcm_s16le', rate=sr)
+    output_stream.layout = 'mono'
+
+    for frame in container.decode(audio=0):
+        frame.pts = None
+        resampled_frames = resampler.resample(frame)
+        if resampled_frames is not None:
+            for resampled_frame in resampled_frames:
+                for packet in output_stream.encode(resampled_frame):
+                    output_container.mux(packet)
+
+    for packet in output_stream.encode(None):
+        output_container.mux(packet)
+
+    output_container.close()
+    return resampled_file

whisper_live/vad.py

@@ -1,18 +1,15 @@
-# original: https://github.com/snakers4/silero-vad/blob/master/utils_vad.py
-
 import os
 import subprocess
 import torch
 import numpy as np
 import onnxruntime
+import warnings
 
 
 class VoiceActivityDetection():
 
     def __init__(self, force_onnx_cpu=True):
-        print("downloading ONNX model...")
         path = self.download()
-        print("loading session")
 
         opts = onnxruntime.SessionOptions()
         opts.log_severity_level = 3
@@ -20,15 +17,17 @@ class VoiceActivityDetection():
         opts.inter_op_num_threads = 1
         opts.intra_op_num_threads = 1
 
-        print("loading onnx model")
         if force_onnx_cpu and 'CPUExecutionProvider' in onnxruntime.get_available_providers():
             self.session = onnxruntime.InferenceSession(path, providers=['CPUExecutionProvider'], sess_options=opts)
         else:
             self.session = onnxruntime.InferenceSession(path, providers=['CUDAExecutionProvider'], sess_options=opts)
 
-        print("reset states")
         self.reset_states()
-        self.sample_rates = [8000, 16000]
+        if '16k' in path:
+            warnings.warn('This model support only 16000 sampling rate!')
+            self.sample_rates = [16000]
+        else:
+            self.sample_rates = [8000, 16000]
 
     def _validate_input(self, x, sr: int):
         if x.dim() == 1:
@@ -43,22 +42,27 @@ class VoiceActivityDetection():
 
         if sr not in self.sample_rates:
             raise ValueError(f"Supported sampling rates: {self.sample_rates} (or multiply of 16000)")
-
         if sr / x.shape[1] > 31.25:
             raise ValueError("Input audio chunk is too short")
 
         return x, sr
 
     def reset_states(self, batch_size=1):
-        self._h = np.zeros((2, batch_size, 64)).astype('float32')
-        self._c = np.zeros((2, batch_size, 64)).astype('float32')
+        self._state = torch.zeros((2, batch_size, 128)).float()
+        self._context = torch.zeros(0)
         self._last_sr = 0
         self._last_batch_size = 0
 
     def __call__(self, x, sr: int):
 
         x, sr = self._validate_input(x, sr)
+        num_samples = 512 if sr == 16000 else 256
+
+        if x.shape[-1] != num_samples:
+            raise ValueError(f"Provided number of samples is {x.shape[-1]} (Supported values: 256 for 8000 sample rate, 512 for 16000)")
+
         batch_size = x.shape[0]
+        context_size = 64 if sr == 16000 else 32
 
         if not self._last_batch_size:
             self.reset_states(batch_size)
@@ -67,28 +71,35 @@ class VoiceActivityDetection():
         if (self._last_batch_size) and (self._last_batch_size != batch_size):
             self.reset_states(batch_size)
 
+        if not len(self._context):
+            self._context = torch.zeros(batch_size, context_size)
+
+        x = torch.cat([self._context, x], dim=1)
         if sr in [8000, 16000]:
-            ort_inputs = {'input': x.numpy(), 'h': self._h, 'c': self._c, 'sr': np.array(sr, dtype='int64')}
+            ort_inputs = {'input': x.numpy(), 'state': self._state.numpy(), 'sr': np.array(sr, dtype='int64')}
             ort_outs = self.session.run(None, ort_inputs)
-            out, self._h, self._c = ort_outs
+            out, state = ort_outs
+            self._state = torch.from_numpy(state)
         else:
             raise ValueError()
 
+        self._context = x[..., -context_size:]
         self._last_sr = sr
         self._last_batch_size = batch_size
 
-        out = torch.tensor(out)
+        out = torch.from_numpy(out)
         return out
 
-    def audio_forward(self, x, sr: int, num_samples: int = 512):
+    def audio_forward(self, x, sr: int):
         outs = []
         x, sr = self._validate_input(x, sr)
+        self.reset_states()
+        num_samples = 512 if sr == 16000 else 256
 
         if x.shape[1] % num_samples:
             pad_num = num_samples - (x.shape[1] % num_samples)
             x = torch.nn.functional.pad(x, (0, pad_num), 'constant', value=0.0)
 
-        self.reset_states(x.shape[0])
         for i in range(0, x.shape[1], num_samples):
             wavs_batch = x[:, i:i+num_samples]
             out_chunk = self.__call__(wavs_batch, sr)
@@ -98,7 +109,7 @@ class VoiceActivityDetection():
         return stacked.cpu()
 
     @staticmethod
-    def download(model_url="https://github.com/snakers4/silero-vad/raw/master/files/silero_vad.onnx"):
+    def download(model_url="https://github.com/snakers4/silero-vad/raw/v5.0/files/silero_vad.onnx"):
         target_dir = os.path.expanduser("~/.cache/whisper-live/")
 
         # Ensure the target directory exists
@@ -110,9 +121,37 @@ class VoiceActivityDetection():
         # Check if the model file already exists
         if not os.path.exists(model_filename):
             # If it doesn't exist, download the model using wget
-            print("Downloading VAD ONNX model...")
             try:
                 subprocess.run(["wget", "-O", model_filename, model_url], check=True)
             except subprocess.CalledProcessError:
                 print("Failed to download the model using wget.")
-        return model_filename
+        return model_filename
+
+
+class VoiceActivityDetector:
+    def __init__(self, threshold=0.5, frame_rate=16000):
+        """
+        Initializes the VoiceActivityDetector with a voice activity detection model and a threshold.
+
+        Args:
+            threshold (float, optional): The probability threshold for detecting voice activity. Defaults to 0.5.
+        """
+        self.model = VoiceActivityDetection()
+        self.threshold = threshold
+        self.frame_rate = frame_rate
+
+    def __call__(self, audio_frame):
+        """
+        Determines if the given audio frame contains speech by comparing the detected speech probability against
+        the threshold.
+
+        Args:
+            audio_frame (np.ndarray): The audio frame to be analyzed for voice activity. It is expected to be a
+                                      NumPy array of audio samples.
+
+        Returns:
+            bool: True if the speech probability exceeds the threshold, indicating the presence of voice activity;
+                  False otherwise.
+        """
+        speech_probs = self.model.audio_forward(torch.from_numpy(audio_frame.copy()), self.frame_rate)[0]
+        return torch.any(speech_probs > self.threshold).item()