Add download srt file option firefox extension

Signed-off-by: makaveli10 <vineet.suryan@collabora.com>

Audio-Transcription-Firefox/README.md

@@ -25,6 +25,7 @@ To capture the audio in the current tab, we used the chrome `tabCapture` API to
 When using the Audio Transcription extension, you have the following options:
  - **Use Collabora Server**: We provide a demo server which runs the whisper small model.
  - **Language**: Select the target language for transcription or translation. You can choose from a variety of languages supported by OpenAI-whisper.
+ - **Download SRT file at Stop Capture**: Select if you want to download the srt file for the session at stop capture.
  - **Task:** Choose the specific task to perform on the audio. You can select either "transcribe" for transcription or "translate" to translate the audio to English.
   - **Model Size**: Select the whisper model size to run the server with.
 

Audio-Transcription-Firefox/audiopreprocessor.js

@@ -0,0 +1,70 @@
+class AudioPreProcessor extends AudioWorkletProcessor {
+    constructor() {
+      super();
+      this.sampleRate = sampleRate || 48000;
+      this.targetSampleRate = 16000;
+      this.inputSamplesNeeded = this.sampleRate * 0.5;
+      this.inputBuffer = new Float32Array(this.inputSamplesNeeded);
+      this.inputWriteOffset = 0;
+    }
+  
+    process(inputs, outputs) {
+      const input = inputs[0]; 
+      const output = outputs[0];
+      if (!input || input.length === 0) {
+        return true;
+      }
+      for (let channel = 0; channel < Math.min(input.length, output.length); channel++) {
+        if (input[channel] && output[channel]) {
+          output[channel].set(input[channel]);
+        }
+      }
+
+      let monoInput;
+      if (input.length === 1) {
+        monoInput = input[0];
+      } else if (input.length > 1) {
+        monoInput = new Float32Array(input[0].length);
+        for (let channel = 0; channel < input.length; channel++) {
+          monoInput.set(input[channel], 0);
+        }
+      }
+
+      if (!monoInput) {
+        return true;
+      }
+  
+      let inputOffset = 0;
+      while (inputOffset < monoInput.length) {
+        const remainingBuffer = this.inputSamplesNeeded - this.inputWriteOffset;
+        const toCopy = Math.min(remainingBuffer, monoInput.length - inputOffset);
+        this.inputBuffer.set(monoInput.subarray(inputOffset, inputOffset + toCopy), this.inputWriteOffset);
+
+        this.inputWriteOffset += toCopy;
+        inputOffset += toCopy;
+
+        if (this.inputWriteOffset === this.inputSamplesNeeded) {
+          const downsampled = this.downsampleTo16kHz(this.inputBuffer);
+          this.port.postMessage(downsampled);
+
+          this.inputWriteOffset = 0;
+        }
+      }
+
+      return true;
+    }
+  
+    downsampleTo16kHz(inputBuffer) {
+      const ratio = this.sampleRate / this.targetSampleRate;
+      const length = Math.floor(inputBuffer.length / ratio); 
+      const result = new Float32Array(length);
+      for (let i = 0; i < length; i++) {
+        const idx = Math.floor(i * ratio);
+        result[i] = inputBuffer[idx];
+      }
+      return result;
+    }
+  }
+  
+  registerProcessor('audiopreprocessor', AudioPreProcessor);
+  

Audio-Transcription-Firefox/content.js

@@ -1,144 +1,162 @@
 let socket = null;
 let isCapturing = false;
-let mediaStream = null;
 let audioContext = null;
-let scriptProcessor = null;
 let language = null;
-
 let isPaused = false;
+let preNode = null;
+let allSegments = [];
+let lastIncompleteSegment = null;
 
-const mediaElements = document.querySelectorAll('video, audio');
+function formatTime(seconds) {
-mediaElements.forEach((mediaElement) => {
+  const date = new Date(seconds * 1000);
-  mediaElement.addEventListener('play', handlePlaybackStateChange);
+  const hh = String(date.getUTCHours()).padStart(2, '0');
-  mediaElement.addEventListener('pause', handlePlaybackStateChange);
+  const mm = String(date.getUTCMinutes()).padStart(2, '0');
-});
+  const ss = String(date.getUTCSeconds()).padStart(2, '0');
-
+  const mmm = String(date.getUTCMilliseconds()).padStart(3, '0');
-
+  return `${hh}:${mm}:${ss},${mmm}`;
-function handlePlaybackStateChange(event) {
-  isPaused = event.target.paused;
 }
 
+function generateSRT() {
+    return allSegments
+    .map((seg, i) => {
+    const start = formatTime(seg.start);
+    const end   = formatTime(seg.end);
+    const text  = seg.text.trim().replace(/[\r\n]+/g, ' ');
+    return `${i + 1}\n${start} --> ${end}\n${text}`;
+    })
+    .join('\n\n');
+}
+
+function downloadSRT() {
+    const srtBlob = new Blob([generateSRT()], { type: 'text/srt;charset=utf-8' });
+    const url = URL.createObjectURL(srtBlob);
+    const a = document.createElement('a');
+    a.href = url;
+    a.download = 'captions.srt';
+    a.style.display = 'none';
+    document.body.appendChild(a);
+    a.click();
+    URL.revokeObjectURL(url);
+    document.body.removeChild(a);
+}
+
+
 function generateUUID() {
   let dt = new Date().getTime();
-  const uuid = 'xxxxxxxx-xxxx-4xxx-yxxx-xxxxxxxxxxxx'.replace(/[xy]/g, function(c) {
+  return 'xxxxxxxx-xxxx-4xxx-yxxx-xxxxxxxxxxxx'.replace(/[xy]/g, c => {
     const r = (dt + Math.random() * 16) % 16 | 0;
     dt = Math.floor(dt / 16);
     return (c === 'x' ? r : (r & 0x3 | 0x8)).toString(16);
   });
-  return uuid;
 }
 
 
-/**
+document.querySelectorAll('video, audio').forEach(el => {
- * Resamples the audio data to a target sample rate of 16kHz.
+  el.addEventListener('play', () => { isPaused = false; });
- * @param {Array|ArrayBuffer|TypedArray} audioData - The input audio data.
+  el.addEventListener('pause', () => { isPaused = true; });
- * @param {number} [origSampleRate=44100] - The original sample rate of the audio data.
+});
- * @returns {Float32Array} The resampled audio data at 16kHz.
- */
-function resampleTo16kHZ(audioData, origSampleRate = 44100) {
-  // Convert the audio data to a Float32Array
-  const data = new Float32Array(audioData);
 
-  // Calculate the desired length of the resampled data
-  const targetLength = Math.round(data.length * (16000 / origSampleRate));
 
-  // Create a new Float32Array for the resampled data
+function setupMessageHandler() {
-  const resampledData = new Float32Array(targetLength);
+  if (preNode) {
+    preNode.port.onmessage = e => {
+      const audio16k = e.data; 
+      if (isCapturing && socket && socket.readyState === WebSocket.OPEN && !isPaused) {
+        socket.send(audio16k);
+      }
+    };
+  }
+}
 
-  // Calculate the spring factor and initialize the first and last values
-  const springFactor = (data.length - 1) / (targetLength - 1);
-  resampledData[0] = data[0];
-  resampledData[targetLength - 1] = data[data.length - 1];
 
-  // Resample the audio data
+const WORKLET_URL = browser.runtime.getURL('audiopreprocessor.js');
-  for (let i = 1; i < targetLength - 1; i++) {
+
-    const index = i * springFactor;
+async function initAudioWorklet() {
-    const leftIndex = Math.floor(index).toFixed();
+  if (audioContext && preNode) {
-    const rightIndex = Math.ceil(index).toFixed();
+    setupMessageHandler();
-    const fraction = index - leftIndex;
+    return;
-    resampledData[i] = data[leftIndex] + (data[rightIndex] - data[leftIndex]) * fraction;
+  }
+  audioContext = new AudioContext();
+  await audioContext.audioWorklet.addModule(WORKLET_URL);
+
+  preNode = new AudioWorkletNode(audioContext, 'audiopreprocessor');
+  document.querySelectorAll('audio, video').forEach(el => {
+    let src;
+    try {
+      src = audioContext.createMediaElementSource(el);
+    } catch(e) {
+      console.warn('Could not create MediaElementSource for', el, e);
+      return;
+    }
+    src.connect(preNode);
+    src.connect(audioContext.destination);
+  });
+
+  preNode.connect(audioContext.destination);
+
+  setupMessageHandler();
+}
+
+async function startRecording(data) {
+  if (!audioContext) {
+    await initAudioWorklet();
   }
 
-  // Return the resampled data
+  const uid = generateUUID();
-  return resampledData;
+  socket = new WebSocket(`ws://${data.host}:${data.port}/`);
+  language = data.language;
+
+  socket.onopen = () => {
+    socket.send(JSON.stringify({
+      uid,
+      language: data.language,
+      task: data.task,
+      model: data.modelSize,
+      use_vad: data.useVad
+    }));
+  };
+
+  let serverReady = false;
+  socket.onmessage = async event => {
+    const msg = JSON.parse(event.data);
+    if (msg.uid !== uid) return;
+
+    if (msg.status === 'WAIT') {
+      await browser.runtime.sendMessage({ action: 'showPopup', data: msg.message });
+      return;
+    }
+    if (!serverReady && msg.message === 'SERVER_READY') {
+      serverReady = true;
+      return;
+    }
+    if (!language && msg.language) {
+      language = msg.language;
+      await browser.runtime.sendMessage({ action: 'updateSelectedLanguage', data: language });
+      return;
+    }
+    if (msg.message === 'DISCONNECT') {
+      await browser.runtime.sendMessage({ action: 'toggleCaptureButtons' });
+      return;
+    }
+    if (msg.segments) {
+      await browser.runtime.sendMessage({ action: 'transcript', data: {data: event.data, saveCaption: data.saveCaption} });
+    }
+  };
+
+  isCapturing = true;
 }
 
-function startRecording(data) {
+function stopRecording() {
-    socket = new WebSocket(`ws://${data.host}:${data.port}/`);
+  isCapturing = false;
-    language = data.language;
+  if (socket) {
+    socket.close();
+    socket = null;
+  }
 
-    const uuid = generateUUID();
+  remove_element();
-    socket.onopen = function(e) { 
-      socket.send(
-        JSON.stringify({
-            uid: uuid,
-            language: data.language,
-            task: data.task,
-            model: data.modelSize,
-            use_vad: data.useVad
-        })
-      );
-    };
-
-    let isServerReady = false;
-    socket.onmessage = async (event) => {
-      const data = JSON.parse(event.data);
-      if (data["uid"] !== uuid)
-        return;
-      
-      if (data["status"] === "WAIT"){
-        await browser.runtime.sendMessage({ action: "showPopup", data: data["message"] })
-        return;
-      }
-      
-      if (!isServerReady && data["message"] === "SERVER_READY"){
-        isServerReady = true;
-        return;
-      }
-
-      if (language === null ){
-        language = data["language"];
-        await browser.runtime.sendMessage({ action: "updateSelectedLanguage", data: language })      
-        return
-      }
-
-      if (data["message"] === "DISCONNECT"){
-        await browser.runtime.sendMessage({ action: "toggleCaptureButtons", data: false })        
-        return
-      }
-
-      await browser.runtime.sendMessage({ action: "transcript", data: event.data })
-          .catch(function(error) {
-            console.error("Error sending message:", error);
-          });
-    };
-
-  // Access the audio stream from the current tab
-  navigator.mediaDevices.getUserMedia({ audio: true })
-    .then(function(stream) {
-      // Create a new MediaRecorder instance
-      const audioDataCache = [];
-      audioContext = new AudioContext();
-      mediaStream = audioContext.createMediaStreamSource(stream);
-      recorder = audioContext.createScriptProcessor(4096, 1, 1);
-
-      recorder.onaudioprocess = async (event) => {
-        if (!audioContext || !isCapturing || !isServerReady || isPaused) return;
-
-        const inputData = event.inputBuffer.getChannelData(0);
-        const audioData16kHz = resampleTo16kHZ(inputData, audioContext.sampleRate);
-
-        audioDataCache.push(inputData);
-        
-        socket.send(audioData16kHz);
-      };
-
-      // Prevent page mute
-      mediaStream.connect(recorder);
-      recorder.connect(audioContext.destination);
-    })
 }
 
+
 var elem_container = null;
 var elem_text = null;
 
@@ -308,6 +326,8 @@ function remove_element() {
 
 browser.runtime.onMessage.addListener((request, sender, sendResponse) => {
   const { action, data } = request;
+  const saveCaption = data.saveCaption || false;
+
   if (action === "startCapture") {
       isCapturing = true;
       startRecording(data);
@@ -318,12 +338,20 @@ browser.runtime.onMessage.addListener((request, sender, sendResponse) => {
         socket.close();
         socket = null;
     }
-    
+
-    if (audioContext) {
+
-        audioContext.close();
+    if (saveCaption === true) {
-        audioContext = null;
+      if (lastIncompleteSegment && lastIncompleteSegment.text && lastIncompleteSegment.text.trim() !== "") {
-        mediaStream = null;
+          if (allSegments.length === 0 || parseFloat(lastIncompleteSegment.start) >= parseFloat(allSegments[allSegments.length - 1].end)) {
-        recorder = null;
+              allSegments.push({
+                  start: lastIncompleteSegment.start,
+                  end: lastIncompleteSegment.end,
+                  text: lastIncompleteSegment.text
+              });
+          }
+      }
+  
+      downloadSRT();
     }
 
     remove_element();
@@ -337,8 +365,25 @@ browser.runtime.onMessage.addListener((request, sender, sendResponse) => {
   } else if (action === "show_transcript"){
     if (!isCapturing) return;
     init_element();    
-    message = JSON.parse(data);
+    message = JSON.parse(data.data);
     message = message["segments"];
+
+    if (saveCaption === true) {
+      message.forEach(seg => {
+          if (seg.completed === true && 
+              (allSegments.length === 0 || parseFloat(seg.start) >= parseFloat(allSegments[allSegments.length - 1].end))) {
+              allSegments.push({
+                  start: seg.start,
+                  end: seg.end,
+                  text: seg.text
+              });
+              
+              lastIncompleteSegment = null;
+          } else if (seg.completed !== true) {
+              lastIncompleteSegment = seg;
+          }
+      });
+    }
     
     var text = '';
     for (var i = 0; i < message.length; i++) {

Audio-Transcription-Firefox/manifest.json

@@ -8,6 +8,9 @@
       "activeTab",
       "<all_urls>"
     ],
+    "web_accessible_resources": [
+      "audiopreprocessor.js"
+    ],
     "background": {
       "scripts": ["background.js"],
       "persistent": false

Audio-Transcription-Firefox/popup.html

@@ -19,6 +19,10 @@
     <input type="checkbox" id="useVadCheckbox">
     <label for="useVadCheckbox">Use Voice Activity Detection</label>
   </div>
+  <div class="checkbox-container">
+    <input type="checkbox" id="saveCaptionCheckbox">
+    <label for="saveCaption">Download SRT file at Stop Capture</label>
+  </div>
   <textarea id="waitTextBox" style="display: none;"></textarea>
   <div class="dropdown-container">
     <label for="languageDropdown">Select Language:</label>

Audio-Transcription-Firefox/popup.js

@@ -4,6 +4,7 @@ document.addEventListener("DOMContentLoaded", function() {
 
   const useServerCheckbox = document.getElementById("useServerCheckbox");
   const useVadCheckbox = document.getElementById("useVadCheckbox");
+  const saveCaptionCheckbox = document.getElementById("saveCaptionCheckbox");
   const languageDropdown = document.getElementById('languageDropdown');
   const taskDropdown = document.getElementById('taskDropdown');
   const modelSizeDropdown = document.getElementById('modelSizeDropdown');
@@ -41,6 +42,12 @@ document.addEventListener("DOMContentLoaded", function() {
     }
   });
 
+  browser.storage.local.get("saveCaptionState", ({ saveCaptionState }) => {
+    if (saveCaptionState !== undefined) {
+      saveCaptionCheckbox.checked = saveCaptionState;
+    }
+  });
+
   browser.storage.local.get("selectedLanguage", ({ selectedLanguage: storedLanguage }) => {
     if (storedLanguage !== undefined) {
       languageDropdown.value = storedLanguage;
@@ -85,6 +92,7 @@ document.addEventListener("DOMContentLoaded", function() {
               task: selectedTask,
               modelSize: selectedModelSize,
               useVad: useVadCheckbox.checked,
+              saveCaption: saveCaptionCheckbox.checked,
             } 
           });
         toggleCaptureButtons(true);
@@ -101,7 +109,7 @@ document.addEventListener("DOMContentLoaded", function() {
   stopButton.addEventListener("click", function() {
     browser.tabs.query({ active: true, currentWindow: true })
       .then(function(tabs) {
-        browser.tabs.sendMessage(tabs[0].id, { action: "stopCapture" })
+        browser.tabs.sendMessage(tabs[0].id, { action: "stopCapture", data: {saveCaption: saveCaptionCheckbox.checked, } })
           .then(function(response) {
             toggleCaptureButtons(false);
             browser.storage.local.set({ capturingState: { isCapturing: false } })
@@ -124,6 +132,7 @@ document.addEventListener("DOMContentLoaded", function() {
     stopButton.disabled = !isCapturing;
     useServerCheckbox.disabled = isCapturing;
     useVadCheckbox.disabled = isCapturing;
+    saveCaptionCheckbox.disabled = isCapturing;
     modelSizeDropdown.disabled = isCapturing;
     languageDropdown.disabled = isCapturing;
     taskDropdown.disabled = isCapturing; 
@@ -142,6 +151,11 @@ document.addEventListener("DOMContentLoaded", function() {
     browser.storage.local.set({ useVadState });
   });
 
+  saveCaptionCheckbox.addEventListener("change", () => {
+    const saveCaptionState = saveCaptionCheckbox.checked;
+    browser.storage.local.set({ saveCaptionState });
+  });
+
   languageDropdown.addEventListener('change', function() {
     if (languageDropdown.value === "") {
       selectedLanguage = null;