cjones/WhisperLive
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge pull request #1 from makaveli10/main

Browse Source 
Initial implementation with faster-whisper.
This commit is contained in:
Marcus Edel
2023-05-05 15:14:37 -04:00
committed by GitHub
parent 72928149ac 28120daad4
commit 396eb78992
7 changed files with 1599 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
+60
View File
@@ -1,2 +1,62 @@
# whisper-live
A nearly-live implementation of OpenAI's Whisper.
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.
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 API and improves the accuracy of the transcription output.
## Installation
- Install PyAudio and ffmpeg
```bash
bash setup.sh
```
- To install client requirements
```bash
pip install -r requirements/client.txt
```
- To install server requirements
```bash
pip install -r requirements/server.txt
```
## Getting Started
- Run the server
```bash
python server.py
```
- On the client side
- To transcribe an audio file:
```bash
python client.py --audio "audio.wav"
```
- To transcribe from microphone:
```bash
python client.py
```
## Citations
```bibtex
@article{Whisper
title = {Robust Speech Recognition via Large-Scale Weak Supervision},
url = {https://arxiv.org/abs/2212.04356},
author = {Radford, Alec and Kim, Jong Wook and Xu, Tao and Brockman, Greg and McLeavey, Christine and Sutskever, Ilya},
publisher = {arXiv},
year = {2022},
}
```
```bibtex
@misc{Silero VAD,
author = {Silero Team},
title = {Silero VAD: pre-trained enterprise-grade Voice Activity Detector (VAD), Number Detector and Language Classifier},
year = {2021},
publisher = {GitHub},
journal = {GitHub repository},
howpublished = {\url{https://github.com/snakers4/silero-vad}},
commit = {insert_some_commit_here},
email = {hello@silero.ai}
}
client.py
+320
View File
@@ -0,0 +1,320 @@
import io
import os
import argparse
import wave
import uuid
import hashlib
import base64
import time
import numpy as np
import scipy
import ffmpeg
import torch
import socket, pickle, pyaudio, struct
import threading
import textwrap
import json
import torchaudio
from dataclasses import dataclass
CHUNK = 1024
FORMAT = pyaudio.paInt16
CHANNELS = 1
RATE = 16000
RECORD_SECONDS = 60000
all_segments = []
@dataclass(frozen=True)
class Constants:
ACK = b"acknowledged"
RECORDING_OVER = b"audio_data_over"
RECEIVED_AUDIO_FILE = b"audio_file_sent"
RECEIVING_AUDIO_FILE = b"sending_audio_file"
class Client:
def __init__(self, topic=None, host=None, port=None):
self.timestamp_offset = 0.0
self.audio_bytes = None
self.p = pyaudio.PyAudio()
self.payload_size = struct.calcsize("Q")
self.stream = self.p.open(format=FORMAT,
channels=CHANNELS,
rate=RATE,
input=True,
frames_per_buffer=CHUNK)
print(self.p.get_sample_size(FORMAT))
self.client_socket = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
host_ip = 'localhost' if host is None else host
port = 5901 if port is None else port
socket_address = (host_ip, port)
self.client_socket.connect(socket_address)
print("CLIENT CONNECTED TO", socket_address)
# voice activity detection model
self.vad_model, _ = torch.hub.load(repo_or_dir='snakers4/silero-vad',
model='silero_vad',
force_reload=True,
onnx=True)
self.window_size = 1024
self.vad_threshold = 0.4
# subscribing to the correct topic
if topic is not None:
self.topic = topic
else:
self.topic = self.get_mac_address().decode()
self.frames = b""
data = b""
while True:
while len(data) < self.payload_size:
packet = self.client_socket.recv(4*1024) #4K
if not packet: break
data+=packet
packed_msg_size = data[:self.payload_size]
data = data[self.payload_size:]
try:
msg_size = struct.unpack("Q",packed_msg_size)[0]
except struct.error:
break
while len(data) < msg_size:
data += self.client_socket.recv(4*1024)
frame_data = data[:msg_size]
frame_data = pickle.loads(frame_data)
if Constants.ACK in frame_data:
print("Server is ready. Sending audio ...")
break
print("* recording")
def send_packet_to_server(self, message):
a = pickle.dumps(message)
message = struct.pack("Q",len(a))+a
self.client_socket.sendall(message)
def get_mac_address(self):
mac = hex(uuid.getnode())
hasher = hashlib.sha1(mac.encode())
return base64.urlsafe_b64encode(hasher.digest()[:5])
@staticmethod
def bytes_to_audio_tensor(audio_bytes):
bytes_io = io.BytesIO()
raw_data = np.frombuffer(
buffer=audio_bytes, dtype=np.int16
)
scipy.io.wavfile.write(bytes_io, RATE, raw_data)
audio, _ = torchaudio.load(bytes_io)
return audio.squeeze(0)
def play_file(self, filename):
# read audio and create pyaudio stream
self.wf = wave.open(filename, 'rb')
self.stream = self.p.open(format=self.p.get_format_from_width(self.wf.getsampwidth()),
channels=self.wf.getnchannels(),
rate=self.wf.getframerate(),
input=True,
output=True,
frames_per_buffer=CHUNK)
try:
while True:
data = self.wf.readframes(CHUNK)
if data==b'': break
# voice activity detection
chunk_tensor = Client.bytes_to_audio_tensor(data)
try:
speech_prob = self.vad_model(chunk_tensor, RATE).item()
except ValueError:
break # input audio chunk is too short
if speech_prob > self.vad_threshold:
data_dict = {
"topic": self.topic,
"audio": data
}
self.send_packet_to_server(data_dict)
self.stream.write(data)
self.wf.close()
self.stream.close()
# let the server know that we're done
data = Constants.RECORDING_OVER
self.send_packet_to_server(data)
with open("results.json", "w") as f:
json_dict = json.dumps(all_segments, indent=2)
f.write(json_dict)
except KeyboardInterrupt:
# write all segments to a file
with open("results.json", "w") as f:
json_dict = json.dumps(all_segments, indent=2)
f.write(json_dict)
def get_client_socket(self):
return self.client_socket
def write_audio_frames_to_file(self, frames, file_name):
wf = wave.open(file_name, 'wb')
wf.setnchannels(CHANNELS)
wf.setsampwidth(2)
wf.setframerate(RATE)
wf.writeframes(frames)
wf.close()
def record(self, out_file="output_recording.wav"):
n_audio_file = 0
# create dir for saving audio chunks
if not os.path.exists("chunks"):
os.makedirs("chunks", exist_ok=True)
try:
for _ in range(0, int(RATE / CHUNK * RECORD_SECONDS)):
data = self.stream.read(CHUNK)
self.frames += data
# voice activity detection
chunk_tensor = Client.bytes_to_audio_tensor(data)
speech_prob = self.vad_model(chunk_tensor, RATE).item()
if speech_prob > self.vad_threshold:
data_dict = {
"topic": self.topic,
"audio": data
}
self.send_packet_to_server(data_dict)
# save frames if more than a minute
if len(self.frames) > 60*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""
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()
# let the server know that we're done
data = Constants.RECORDING_OVER
self.send_packet_to_server(data)
# combine all the audio files
self.write_output_recording(n_audio_file, out_file)
# write all segments to a file
with open("results.json", "w") as f:
json_dict = json.dumps(all_segments, indent=2)
f.write(json_dict)
def write_output_recording(self, n_audio_file, out_file):
input_files = [f"chunks/{i}.wav" for i in range(n_audio_file) if os.path.exists(f"chunks/{i}.wav")]
wf = wave.open(out_file, 'wb')
wf.setnchannels(CHANNELS)
wf.setsampwidth(2)
wf.setframerate(RATE)
for in_file in input_files:
w = wave.open(in_file, 'rb')
while True:
data = w.readframes(CHUNK)
if data==b'': break
wf.writeframes(data)
w.close()
# remove this file
os.remove(in_file)
wf.close()
def recieve_response(client_socket):
data = b""
payload_size = struct.calcsize("Q")
while True:
while len(data) < payload_size:
packet = client_socket.recv(4*1024) # 4K
if not packet: break
data+=packet
packed_msg_size = data[:payload_size]
data = data[payload_size:]
try:
msg_size = struct.unpack("Q",packed_msg_size)[0]
except struct.error:
break
while len(data) < msg_size:
data += client_socket.recv(4*1024)
frame_data = data[:msg_size]
data = data[msg_size:]
response = pickle.loads(frame_data)
if response is not None and isinstance(response, dict):
os.system('clear')
text = response['text']
segments = response['segments']
if len(segments):
for seg in segments:
all_segments.append(seg)
wrapper = textwrap.TextWrapper(width=50)
word_list = wrapper.wrap(text=text)
# Print each line.
for element in word_list:
print(element)
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
Parameters
----------
file: str
The audio file to open
sr: int
The sample rate to resample the audio if necessary
"""
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
if __name__=="__main__":
parser = argparse.ArgumentParser()
parser.add_argument('--audio', type=str, help='audio file to transcribe')
parser.add_argument('--topic', default=None, type=str, help='topic to subscribe for results')
parser.add_argument('--host', default=None, type=str, help='server address to connect to')
parser.add_argument('--port', default=None, type=str, help='server port to connect to')
opt = parser.parse_args()
c = Client(topic=opt.topic, host=opt.host, port=opt.port)
while True:
if c.get_client_socket() is not None:
break
client_socket = c.get_client_socket()
t2 = threading.Thread(target=recieve_response, args=(client_socket, ))
t2.start()
if opt.audio is not None:
resampled_file = resample(opt.audio)
c.play_file(resampled_file)
else:
c.record()
t2.join()
requirements/client.txt
+5
View File
@@ -0,0 +1,5 @@
PyAudio
ffmpeg-python
scipy
torch==1.12.1
torchaudio==0.12.1
requirements/server.txt
+5
View File
@@ -0,0 +1,5 @@
PyAudio
faster-whisper==0.4.1
paho-mqtt
--extra-index-url https://download.pytorch.org/whl/cu113
torch==1.12.1
server.py
+331
View File
@@ -0,0 +1,331 @@
import socket, pickle, struct, time, pyaudio
import threading
import os
import wave
import textwrap
from collections import deque
from dataclasses import dataclass
import torch
import numpy as np
import paho.mqtt.client as mqtt
from transcriber import WhisperModel
def on_connect(mqttc, obj, flags, rc):
pass
def on_message(mqttc, obj, msg):
pass
def on_publish(mqttc, obj, mid):
pass
def on_subscribe(mqttc, obj, mid, granted_qos):
pass
def on_log(mqttc, obj, level, string):
pass
@dataclass(frozen=True)
class Constants:
AUDIO_OVER = b"audio_data_over"
ACK = b"acknowledged"
SENDING_FILE = b"sending_audio_file"
FILE_SENT = b"audio_file_sent"
class ServeClient:
CHUNK = 1024
FORMAT = pyaudio.paInt16
CHANNELS = 1
RATE = 16000
def __init__(self, client_socket, device=None, verbose=True):
self.payload_size = struct.calcsize("Q")
self.data = b""
self.frames = b""
self.frames_np = None
self.transcriber = WhisperModel("medium.en", device="cuda", compute_type="float16")
self.timestamp_offset = 0.0
self.frames_offset = 0.0
self.text = []
self.current_out = ''
self.prev_out = ''
self.t_start=None
self.client_socket = client_socket
self.verbose = verbose
self.exit = False
self.same_output_threshold = 0
self.show_prev_out_thresh = 5 # if pause(no output from whisper) show previous output for 5 seconds
self.add_pause_thresh = 3 # add a blank to segment list as a pause(no speech) for 3 seconds
# text formatting
self.wrapper = textwrap.TextWrapper(width=50)
self.pick_previous_segments = 2
# setup mqtt
self.topic = None
self.mqttc = mqtt.Client()
self.mqttc.on_message = on_message
self.mqttc.on_connect = on_connect
self.mqttc.on_publish = on_publish
self.mqttc.on_subscribe = on_subscribe
self.mqttc = mqtt.Client()
self.mqttc.connect("mqtt.kurg.org", 1883, 60)
self.mqttc.loop_start()
# send response to client; server is ready
self.send_response_to_client(Constants.ACK)
# threading
self.recv_thread = threading.Thread(target=self.recv_audio)
self.trans_thread = threading.Thread(target=self.speech_to_text)
self.recv_thread.start()
self.trans_thread.start()
def recv_audio(self):
"""
Receive audio chunks from client in an infinite loop.
"""
if self.client_socket:
try:
while True:
while len(self.data) < self.payload_size:
packet = self.client_socket.recv(4*1024) # 4K
if not packet: break
self.data+=packet
packed_msg_size = self.data[:self.payload_size]
self.data = self.data[self.payload_size:]
msg_size = struct.unpack("Q",packed_msg_size)[0]
while len(self.data) < msg_size:
self.data += self.client_socket.recv(4*1024)
frame_data = self.data[:msg_size]
self.data = self.data[msg_size:]
frame_data = pickle.loads(frame_data)
if self.topic is None:
self.topic = frame_data["topic"]
frame = frame_data["audio"]
# client says audio over
if Constants.AUDIO_OVER in frame:
break
frame_np = np.frombuffer(frame, dtype=np.int16)
if self.frames_np is not None and self.frames_np.shape[0] > 60*self.RATE:
self.frames_offset += 45.0
self.frames_np = self.frames_np[int(45*self.RATE):]
if self.frames_np is None:
self.frames_np = frame_np.copy()
else:
self.frames_np = np.concatenate((self.frames_np, frame_np), axis=0)
# set frames np to None so to stop translation for this client
self.frames_np = None
self.exit = True
except Exception as e:
if self.verbose: print(f"[ERROR]: {e}")
self.exit = True
def send_response_to_client(self, message):
"""
Send serialized response to client.
"""
a = pickle.dumps(message)
message = struct.pack("Q",len(a))+a
self.client_socket.sendall(message)
def fill_output(self, output):
"""
Format output with current and previous complete segments
into two lines of 50 characters.
Args:
output(str): current incomplete segment
Returns:
transcription wrapped in two lines
"""
text = ''
pick_prev = min(len(self.text), self.pick_previous_segments)
for seg in self.text[-pick_prev:]:
# discard everything before a 3 second pause
if seg == '':
text = ''
else:
text += seg
wrapped = self.wrapper.wrap(
text="".join(text + output))[-2:]
return " ".join(wrapped)
def speech_to_text(self):
"""
Process audio stream in an infinite loop.
"""
while True:
if self.exit:
self.mqttc.disconnect()
self.client_socket.close()
self.transcriber.destroy()
break
if self.frames_np is None: continue
# clip audio if the current chunk exceeds 25 seconds, this basically implies that
# no valid segment for the last 25 seconds from whisper
if self.frames_np[int((self.timestamp_offset - self.frames_offset)*self.RATE):].shape[0] > 25 * self.RATE:
duration = self.frames_np.shape[0] / self.RATE
self.timestamp_offset = self.frames_offset + duration - 5
# add 200 ms from the last chunk if available
if len(self.text) and self.frames_np[:-int((self.timestamp_offset - self.frames_offset)*self.RATE)].shape[0]:
samples_take = max(0, (self.timestamp_offset - self.frames_offset)*self.RATE - 0.2*self.RATE)
else:
samples_take = max(0, (self.timestamp_offset - self.frames_offset)*self.RATE)
input_bytes = self.frames_np[int(samples_take):].copy()
duration = input_bytes.shape[0] / self.RATE
if duration<1.0: continue
try:
input_sample = input_bytes.astype(np.float32) / 32768.0
# set previous complete segment as initial prompt
if len(self.text) and self.text[-1] != '':
initial_prompt = self.text[-1]
else:
initial_prompt = None
# whisper transcribe with prompt
result = self.transcriber.transcribe(input_sample, initial_prompt=initial_prompt)
if len(result):
self.t_start = None
output, segments = self.update_segments(result, duration)
out_dict = {
'text': output,
'segments': segments
}
if self.topic is not None:
self.mqttc.publish(self.topic, payload=str(out_dict))
self.send_response_to_client(out_dict)
else:
# show previous output if there is pause i.e. no output from whisper
output = ''
if self.t_start is None: self.t_start = time.time()
if time.time() - self.t_start < self.show_prev_out_thresh:
output = self.fill_output('')
# add a blank if there is no speech for 3 seconds
if len(self.text) and self.text[-1] != '':
if time.time() - self.t_start > self.add_pause_thresh:
self.text.append('')
# publish outputs
out_dict = {
'text': output,
'segments': []
}
if self.topic is not None:
self.mqttc.publish(self.topic, payload=str(out_dict))
self.send_response_to_client(out_dict)
except Exception as e:
if self.verbose: print(f"[ERROR]: {e}")
time.sleep(0.01)
def update_segments(self, segments, duration):
"""
Processes the segments from whisper. Appends all the segments to the list
except for the last segment assuming that it is incomplete.
Args:
segments(dict) : dictionary of segments as returned by whisper
duration(float): duration of the current chunk
Returns:
transcription for the current chunk
"""
offset = None
transcript = []
self.current_out = ''
# process complete segments
if len(segments) > 1:
for i, s in enumerate(segments[:-1]):
text_ = s.text
self.text.append(text_)
start, end = self.timestamp_offset + s.start, self.timestamp_offset + min(duration, s.end)
transcript.append(
{
'start': start,
'end': end,
'text': text_
}
)
offset = min(duration, s.end)
self.current_out += segments[-1].text
# if same incomplete segment is seen multiple times then update the offset
# and append the segment to the list
if self.current_out.strip() == self.prev_out.strip() and self.current_out != '':
self.same_output_threshold += 1
else:
self.same_output_threshold = 0
if self.same_output_threshold > 5:
if not len(self.text) or self.text[-1].strip().lower()!=self.current_out.strip().lower():
self.text.append(self.current_out)
transcript.append(
{
'start': self.timestamp_offset,
'end': self.timestamp_offset + duration,
'text': self.current_out
}
)
self.current_out = ''
offset = duration
self.same_output_threshold = 0
else:
self.prev_out = self.current_out
# update offset
if offset is not None:
self.timestamp_offset += offset
# format and return output
output = self.current_out
return self.fill_output(output), transcript
if __name__=="__main__":
# create socket
server_socket = socket.socket(socket.AF_INET,socket.SOCK_STREAM)
server_socket.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
host='127.0.0.1'
port=5901
backlog=5
socket_address = (host, port)
print('STARTING SERVER AT',socket_address,'...')
server_socket.bind(socket_address)
server_socket.listen(backlog)
client_sockets = []
device = 0
try:
while True:
client_socket, addr = server_socket.accept()
print('GOT CONNECTION FROM:', addr)
client = ServeClient(client_socket, device=f'cuda:{device}')
client_sockets.append(client_socket)
print("waiting for new connection")
except Exception as e:
print(f"[ERROR main]: {e}")
for sock in client_sockets:
try:
sock.close()
except:
pass
setup.sh
+4
View File
@@ -0,0 +1,4 @@
#! /bin/bash
apt-get install portaudio19-dev -y
apt-get install -y --no-install-recommends ffmpeg
transcriber.py
+874
View File
@@ -0,0 +1,874 @@
# original https://github.com/guillaumekln/faster-whisper/blob/master/faster_whisper/transcribe.py
import itertools
import logging
import os
import zlib
import logging
from typing import BinaryIO, Iterable, List, NamedTuple, Optional, Tuple, Union
import ctranslate2
import numpy as np
import tokenizers
from faster_whisper.audio import decode_audio
from faster_whisper.feature_extractor import FeatureExtractor
from faster_whisper.tokenizer import Tokenizer
from faster_whisper.utils import download_model, format_timestamp
from faster_whisper.vad import (
SpeechTimestampsMap,
collect_chunks,
get_speech_timestamps,
)
# implement logger not available in faster_whisper==0.4.1
def get_logger():
"""Returns the module logger."""
return logging.getLogger("faster_whisper")
class Word(NamedTuple):
start: float
end: float
word: str
probability: float
class Segment(NamedTuple):
start: float
end: float
text: str
words: Optional[List[Word]]
avg_log_prob: float
no_speech_prob: float
class AudioInfo(NamedTuple):
language: str
language_probability: float
duration: float
class TranscriptionOptions(NamedTuple):
beam_size: int
best_of: int
patience: float
length_penalty: float
log_prob_threshold: Optional[float]
no_speech_threshold: Optional[float]
compression_ratio_threshold: Optional[float]
condition_on_previous_text: bool
temperatures: List[float]
initial_prompt: Optional[str]
prefix: Optional[str]
suppress_blank: bool
suppress_tokens: Optional[List[int]]
without_timestamps: bool
max_initial_timestamp: float
word_timestamps: bool
prepend_punctuations: str
append_punctuations: str
class WhisperModel:
def __init__(
self,
model_size_or_path: str,
device: str = "auto",
device_index: Union[int, List[int]] = 0,
compute_type: str = "default",
cpu_threads: int = 0,
num_workers: int = 1,
download_root: Optional[str] = None,
):
"""Initializes the Whisper model.
Args:
model_size_or_path: Size of the model to use (tiny, tiny.en, base, base.en,
small, small.en, medium, medium.en, large-v1, or large-v2) or a path to a converted
model directory. When a size is configured, the converted model is downloaded
from the Hugging Face Hub.
device: Device to use for computation ("cpu", "cuda", "auto").
device_index: Device ID to use.
The model can also be loaded on multiple GPUs by passing a list of IDs
(e.g. [0, 1, 2, 3]). In that case, multiple transcriptions can run in parallel
when transcribe() is called from multiple Python threads (see also num_workers).
compute_type: Type to use for computation.
See https://opennmt.net/CTranslate2/quantization.html.
cpu_threads: Number of threads to use when running on CPU (4 by default).
A non zero value overrides the OMP_NUM_THREADS environment variable.
num_workers: When transcribe() is called from multiple Python threads,
having multiple workers enables true parallelism when running the model
(concurrent calls to self.model.generate() will run in parallel).
This can improve the global throughput at the cost of increased memory usage.
download_root: Directory where the model should be saved. If not set, the model
is saved in the standard Hugging Face cache directory.
"""
self.logger = get_logger()
if os.path.isdir(model_size_or_path):
model_path = model_size_or_path
else:
model_path = download_model(model_size_or_path, download_root)
self.model = ctranslate2.models.Whisper(
model_path,
device=device,
device_index=device_index,
compute_type=compute_type,
intra_threads=cpu_threads,
inter_threads=num_workers,
)
tokenizer_file = os.path.join(model_path, "tokenizer.json")
if os.path.isfile(tokenizer_file):
self.hf_tokenizer = tokenizers.Tokenizer.from_file(tokenizer_file)
else:
self.hf_tokenizer = tokenizers.Tokenizer.from_pretrained(
"openai/whisper-tiny" + ("" if self.model.is_multilingual else ".en")
)
self.feature_extractor = FeatureExtractor()
self.num_samples_per_token = self.feature_extractor.hop_length * 2
self.frames_per_second = (
self.feature_extractor.sampling_rate // self.feature_extractor.hop_length
)
self.tokens_per_second = (
self.feature_extractor.sampling_rate // self.num_samples_per_token
)
self.input_stride = 2
self.time_precision = 0.02
self.max_length = 448
def transcribe(
self,
audio: Union[str, BinaryIO, np.ndarray],
language: Optional[str] = None,
task: str = "transcribe",
beam_size: int = 5,
best_of: int = 5,
patience: float = 1,
length_penalty: float = 1,
temperature: Union[float, List[float], Tuple[float, ...]] = [
0.0,
0.2,
0.4,
0.6,
0.8,
1.0,
],
compression_ratio_threshold: Optional[float] = 2.4,
log_prob_threshold: Optional[float] = -1.0,
no_speech_threshold: Optional[float] = 0.6,
condition_on_previous_text: bool = True,
initial_prompt: Optional[str] = None,
prefix: Optional[str] = None,
suppress_blank: bool = True,
suppress_tokens: Optional[List[int]] = [-1],
without_timestamps: bool = False,
max_initial_timestamp: float = 1.0,
word_timestamps: bool = False,
prepend_punctuations: str = "\"'“¿([{-",
append_punctuations: str = "\"'.。,，!！?？:：”)]}、",
vad_filter: bool = False,
vad_parameters: Optional[dict] = None,
) -> Tuple[Iterable[Segment], AudioInfo]:
"""Transcribes an input file.
Arguments:
audio: Path to the input file (or a file-like object), or the audio waveform.
language: The language spoken in the audio. It should be a language code such
as "en" or "fr". If not set, the language will be detected in the first 30 seconds
of audio.
task: Task to execute (transcribe or translate).
beam_size: Beam size to use for decoding.
best_of: Number of candidates when sampling with non-zero temperature.
patience: Beam search patience factor.
length_penalty: Exponential length penalty constant.
temperature: Temperature for sampling. It can be a tuple of temperatures,
which will be successively used upon failures according to either
`compression_ratio_threshold` or `log_prob_threshold`.
compression_ratio_threshold: If the gzip compression ratio is above this value,
treat as failed.
log_prob_threshold: If the average log probability over sampled tokens is
below this value, treat as failed.
no_speech_threshold: If the no_speech probability is higher than this value AND
the average log probability over sampled tokens is below `log_prob_threshold`,
consider the segment as silent.
condition_on_previous_text: If True, the previous output of the model is provided
as a prompt for the next window; disabling may make the text inconsistent across
windows, but the model becomes less prone to getting stuck in a failure loop,
such as repetition looping or timestamps going out of sync.
initial_prompt: Optional text to provide as a prompt for the first window.
prefix: Optional text to provide as a prefix for the first window.
suppress_blank: Suppress blank outputs at the beginning of the sampling.
suppress_tokens: List of token IDs to suppress. -1 will suppress a default set
of symbols as defined in the model config.json file.
without_timestamps: Only sample text tokens.
max_initial_timestamp: The initial timestamp cannot be later than this.
word_timestamps: Extract word-level timestamps using the cross-attention pattern
and dynamic time warping, and include the timestamps for each word in each segment.
prepend_punctuations: If word_timestamps is True, merge these punctuation symbols
with the next word
append_punctuations: If word_timestamps is True, merge these punctuation symbols
with the previous word
vad_filter: Enable the voice activity detection (VAD) to filter out parts of the audio
without speech. This step is using the Silero VAD model
https://github.com/snakers4/silero-vad.
vad_parameters: Dictionary of Silero VAD parameters (see available parameters and
default values in the function `get_speech_timestamps`).
Returns:
A tuple with:
- a generator over transcribed segments
- an instance of AudioInfo
"""
sampling_rate = self.feature_extractor.sampling_rate
if not isinstance(audio, np.ndarray):
audio = decode_audio(audio, sampling_rate=sampling_rate)
duration = audio.shape[0] / sampling_rate
self.logger.info(
"Processing audio with duration %s", format_timestamp(duration)
)
if vad_filter:
vad_parameters = {} if vad_parameters is None else vad_parameters
speech_chunks = get_speech_timestamps(audio, **vad_parameters)
audio = collect_chunks(audio, speech_chunks)
self.logger.info(
"VAD filter removed %s of audio",
format_timestamp(duration - (audio.shape[0] / sampling_rate)),
)
if self.logger.isEnabledFor(logging.DEBUG):
self.logger.debug(
"VAD filter kept the following audio segments: %s",
", ".join(
"[%s -> %s]"
% (
format_timestamp(chunk["start"] / sampling_rate),
format_timestamp(chunk["end"] / sampling_rate),
)
for chunk in speech_chunks
),
)
else:
speech_chunks = None
features = self.feature_extractor(audio)
encoder_output = None
if language is None:
if not self.model.is_multilingual:
language = "en"
language_probability = 1
else:
segment = features[:, : self.feature_extractor.nb_max_frames]
encoder_output = self.encode(segment)
results = self.model.detect_language(encoder_output)
language_token, language_probability = results[0][0]
language = language_token[2:-2]
self.logger.info(
"Detected language '%s' with probability %.2f",
language,
language_probability,
)
else:
language_probability = 1
tokenizer = Tokenizer(
self.hf_tokenizer,
self.model.is_multilingual,
task=task,
language=language,
)
options = TranscriptionOptions(
beam_size=beam_size,
best_of=best_of,
patience=patience,
length_penalty=length_penalty,
log_prob_threshold=log_prob_threshold,
no_speech_threshold=no_speech_threshold,
compression_ratio_threshold=compression_ratio_threshold,
condition_on_previous_text=condition_on_previous_text,
temperatures=(
temperature if isinstance(temperature, (list, tuple)) else [temperature]
),
initial_prompt=initial_prompt,
prefix=prefix,
suppress_blank=suppress_blank,
suppress_tokens=get_suppressed_tokens(tokenizer, suppress_tokens),
without_timestamps=without_timestamps,
max_initial_timestamp=max_initial_timestamp,
word_timestamps=word_timestamps,
prepend_punctuations=prepend_punctuations,
append_punctuations=append_punctuations,
)
segments = self.generate_segments(features, tokenizer, options, encoder_output)
if speech_chunks:
segments = restore_speech_timestamps(segments, speech_chunks, sampling_rate)
audio_info = AudioInfo(
language=language,
language_probability=language_probability,
duration=duration,
)
return segments
def generate_segments(
self,
features: np.ndarray,
tokenizer: Tokenizer,
options: TranscriptionOptions,
encoder_output: Optional[ctranslate2.StorageView] = None,
) -> Iterable[Segment]:
content_frames = features.shape[-1] - self.feature_extractor.nb_max_frames
seek = 0
all_tokens = []
prompt_reset_since = 0
if options.initial_prompt is not None:
initial_prompt = " " + options.initial_prompt.strip()
initial_prompt_tokens = tokenizer.encode(initial_prompt)
all_tokens.extend(initial_prompt_tokens)
all_segments = []
while seek < content_frames:
time_offset = seek * self.feature_extractor.time_per_frame
segment = features[:, seek : seek + self.feature_extractor.nb_max_frames]
segment_size = min(
self.feature_extractor.nb_max_frames, content_frames - seek
)
segment_duration = segment_size * self.feature_extractor.time_per_frame
if self.logger.isEnabledFor(logging.DEBUG):
self.logger.debug(
"Processing segment at %s", format_timestamp(time_offset)
)
previous_tokens = all_tokens[prompt_reset_since:]
prompt = self.get_prompt(
tokenizer,
previous_tokens,
without_timestamps=options.without_timestamps,
prefix=options.prefix if seek == 0 else None,
)
if encoder_output is None:
encoder_output = self.encode(segment)
result, avg_log_prob, temperature = self.generate_with_fallback(
encoder_output, prompt, tokenizer, options
)
if options.no_speech_threshold is not None:
# no voice activity check
should_skip = result.no_speech_prob > options.no_speech_threshold
if (
options.log_prob_threshold is not None
and avg_log_prob > options.log_prob_threshold
):
# don't skip if the logprob is high enough, despite the no_speech_prob
should_skip = False
if should_skip:
self.logger.debug(
"No speech threshold is met (%f > %f)",
result.no_speech_prob,
options.no_speech_threshold,
)
# fast-forward to the next segment boundary
seek += segment_size
continue
tokens = result.sequences_ids[0]
previous_seek = seek
current_segments = []
single_timestamp_ending = (
len(tokens) >= 2
and tokens[-2] < tokenizer.timestamp_begin
and tokens[-1] >= tokenizer.timestamp_begin
)
consecutive_timestamps = [
i
for i in range(len(tokens))
if i > 0
and tokens[i] >= tokenizer.timestamp_begin
and tokens[i - 1] >= tokenizer.timestamp_begin
]
if len(consecutive_timestamps) > 0:
slices = list(consecutive_timestamps)
if single_timestamp_ending:
slices.append(len(tokens))
last_slice = 0
for current_slice in slices:
sliced_tokens = tokens[last_slice:current_slice]
start_timestamp_position = (
sliced_tokens[0] - tokenizer.timestamp_begin
)
end_timestamp_position = (
sliced_tokens[-1] - tokenizer.timestamp_begin
)
start_time = (
time_offset + start_timestamp_position * self.time_precision
)
end_time = (
time_offset + end_timestamp_position * self.time_precision
)
current_segments.append(
dict(
seek=seek,
start=start_time,
end=end_time,
tokens=sliced_tokens,
)
)
last_slice = current_slice
if single_timestamp_ending:
# single timestamp at the end means no speech after the last timestamp.
seek += segment_size
else:
# otherwise, ignore the unfinished segment and seek to the last timestamp
last_timestamp_position = (
tokens[last_slice - 1] - tokenizer.timestamp_begin
)
seek += last_timestamp_position * self.input_stride
else:
duration = segment_duration
timestamps = [
token for token in tokens if token >= tokenizer.timestamp_begin
]
if len(timestamps) > 0 and timestamps[-1] != tokenizer.timestamp_begin:
last_timestamp_position = timestamps[-1] - tokenizer.timestamp_begin
duration = last_timestamp_position * self.time_precision
current_segments.append(
dict(
seek=seek,
start=time_offset,
end=time_offset + duration,
tokens=tokens,
)
)
seek += segment_size
if not options.condition_on_previous_text or temperature > 0.5:
prompt_reset_since = len(all_tokens)
if options.word_timestamps:
self.add_word_timestamps(
current_segments,
tokenizer,
encoder_output,
segment_size,
options.prepend_punctuations,
options.append_punctuations,
)
word_end_timestamps = [
w["end"] for s in current_segments for w in s["words"]
]
if not single_timestamp_ending and len(word_end_timestamps) > 0:
seek_shift = round(
(word_end_timestamps[-1] - time_offset) * self.frames_per_second
)
if seek_shift > 0:
seek = previous_seek + seek_shift
encoder_output = None
for segment in current_segments:
tokens = segment["tokens"]
text = tokenizer.decode(tokens)
if segment["start"] == segment["end"] or not text.strip():
continue
all_tokens.extend(tokens)
all_segments.append(Segment(
start=segment["start"],
end=segment["end"],
text=text,
words=(
[Word(**word) for word in segment["words"]]
if options.word_timestamps
else None
),
avg_log_prob=avg_log_prob,
no_speech_prob=result.no_speech_prob,
))
return all_segments
def encode(self, features: np.ndarray) -> ctranslate2.StorageView:
# When the model is running on multiple GPUs, the encoder output should be moved
# to the CPU since we don't know which GPU will handle the next job.
to_cpu = self.model.device == "cuda" and len(self.model.device_index) > 1
features = np.expand_dims(features, 0)
features = get_ctranslate2_storage(features)
return self.model.encode(features, to_cpu=to_cpu)
def generate_with_fallback(
self,
encoder_output: ctranslate2.StorageView,
prompt: List[int],
tokenizer: Tokenizer,
options: TranscriptionOptions,
) -> Tuple[ctranslate2.models.WhisperGenerationResult, float, float]:
result = None
avg_log_prob = None
final_temperature = None
max_initial_timestamp_index = int(
round(options.max_initial_timestamp / self.time_precision)
)
for temperature in options.temperatures:
if temperature > 0:
kwargs = {
"beam_size": 1,
"num_hypotheses": options.best_of,
"sampling_topk": 0,
"sampling_temperature": temperature,
}
else:
kwargs = {
"beam_size": options.beam_size,
"patience": options.patience,
}
final_temperature = temperature
result = self.model.generate(
encoder_output,
[prompt],
length_penalty=options.length_penalty,
max_length=self.max_length,
return_scores=True,
return_no_speech_prob=True,
suppress_blank=options.suppress_blank,
suppress_tokens=options.suppress_tokens,
max_initial_timestamp_index=max_initial_timestamp_index,
**kwargs,
)[0]
tokens = result.sequences_ids[0]
# Recover the average log prob from the returned score.
seq_len = len(tokens)
cum_log_prob = result.scores[0] * (seq_len**options.length_penalty)
avg_log_prob = cum_log_prob / (seq_len + 1)
text = tokenizer.decode(tokens).strip()
compression_ratio = get_compression_ratio(text)
needs_fallback = False
if (
options.compression_ratio_threshold is not None
and compression_ratio > options.compression_ratio_threshold
):
needs_fallback = True # too repetitive
self.logger.debug(
"Compression ratio threshold is not met with temperature %.1f (%f > %f)",
temperature,
compression_ratio,
options.compression_ratio_threshold,
)
if (
options.log_prob_threshold is not None
and avg_log_prob < options.log_prob_threshold
):
needs_fallback = True # average log probability is too low
self.logger.debug(
"Log probability threshold is not met with temperature %.1f (%f < %f)",
temperature,
avg_log_prob,
options.log_prob_threshold,
)
if not needs_fallback:
break
return result, avg_log_prob, final_temperature
def get_prompt(
self,
tokenizer: Tokenizer,
previous_tokens: List[int],
without_timestamps: bool = False,
prefix: Optional[str] = None,
) -> List[int]:
prompt = []
if previous_tokens:
prompt.append(tokenizer.sot_prev)
prompt.extend(previous_tokens[-(self.max_length // 2 - 1) :])
prompt.extend(tokenizer.sot_sequence)
if without_timestamps:
prompt.append(tokenizer.no_timestamps)
if prefix:
prefix_tokens = tokenizer.encode(" " + prefix.strip())
if len(prefix_tokens) >= self.max_length // 2:
prefix_tokens = prefix_tokens[: self.max_length // 2 - 1]
prompt.extend(prefix_tokens)
return prompt
def add_word_timestamps(
self,
segments: List[dict],
tokenizer: Tokenizer,
encoder_output: ctranslate2.StorageView,
num_frames: int,
prepend_punctuations: str,
append_punctuations: str,
):
if len(segments) == 0:
return
text_tokens_per_segment = [
[token for token in segment["tokens"] if token < tokenizer.eot]
for segment in segments
]
text_tokens = list(itertools.chain.from_iterable(text_tokens_per_segment))
alignment = self.find_alignment(
tokenizer, text_tokens, encoder_output, num_frames
)
merge_punctuations(alignment, prepend_punctuations, append_punctuations)
time_offset = (
segments[0]["seek"]
* self.feature_extractor.hop_length
/ self.feature_extractor.sampling_rate
)
word_index = 0
for segment, text_tokens in zip(segments, text_tokens_per_segment):
saved_tokens = 0
words = []
while word_index < len(alignment) and saved_tokens < len(text_tokens):
timing = alignment[word_index]
if timing["word"]:
words.append(
dict(
word=timing["word"],
start=round(time_offset + timing["start"], 2),
end=round(time_offset + timing["end"], 2),
probability=timing["probability"],
)
)
saved_tokens += len(timing["tokens"])
word_index += 1
if len(words) > 0:
# adjust the segment-level timestamps based on the word-level timestamps
segment["start"] = words[0]["start"]
segment["end"] = words[-1]["end"]
segment["words"] = words
def find_alignment(
self,
tokenizer: Tokenizer,
text_tokens: List[int],
encoder_output: ctranslate2.StorageView,
num_frames: int,
median_filter_width: int = 7,
) -> List[dict]:
if len(text_tokens) == 0:
return []
result = self.model.align(
encoder_output,
tokenizer.sot_sequence,
[text_tokens],
num_frames,
median_filter_width=median_filter_width,
)[0]
text_token_probs = result.text_token_probs
alignments = result.alignments
text_indices = np.array([pair[0] for pair in alignments])
time_indices = np.array([pair[1] for pair in alignments])
words, word_tokens = tokenizer.split_to_word_tokens(
text_tokens + [tokenizer.eot]
)
word_boundaries = np.pad(np.cumsum([len(t) for t in word_tokens[:-1]]), (1, 0))
jumps = np.pad(np.diff(text_indices), (1, 0), constant_values=1).astype(bool)
jump_times = time_indices[jumps] / self.tokens_per_second
start_times = jump_times[word_boundaries[:-1]]
end_times = jump_times[word_boundaries[1:]]
word_probabilities = [
np.mean(text_token_probs[i:j])
for i, j in zip(word_boundaries[:-1], word_boundaries[1:])
]
# hack: ensure the first and second word is not longer than twice the median word duration.
# a better segmentation algorithm based on VAD should be able to replace this.
word_durations = end_times - start_times
word_durations = word_durations[word_durations.nonzero()]
if len(word_durations) > 0:
median_duration = np.median(word_durations)
max_duration = median_duration * 2
if len(word_durations) >= 2 and word_durations[1] > max_duration:
boundary = max(end_times[2] / 2, end_times[2] - max_duration)
end_times[0] = start_times[1] = boundary
if (
len(word_durations) >= 1
and end_times[0] - start_times[0] > max_duration
):
start_times[0] = max(0, end_times[0] - max_duration)
return [
dict(
word=word, tokens=tokens, start=start, end=end, probability=probability
)
for word, tokens, start, end, probability in zip(
words, word_tokens, start_times, end_times, word_probabilities
)
]
def destroy(self):
del self.model
def restore_speech_timestamps(
segments: Iterable[Segment],
speech_chunks: List[dict],
sampling_rate: int,
) -> Iterable[Segment]:
ts_map = SpeechTimestampsMap(speech_chunks, sampling_rate)
for segment in segments:
if segment.words:
words = []
for word in segment.words:
# Ensure the word start and end times are resolved to the same chunk.
chunk_index = ts_map.get_chunk_index(word.start)
word = word._replace(
start=ts_map.get_original_time(word.start, chunk_index),
end=ts_map.get_original_time(word.end, chunk_index),
)
words.append(word)
segment = segment._replace(
start=words[0].start,
end=words[-1].end,
words=words,
)
else:
segment = segment._replace(
start=ts_map.get_original_time(segment.start),
end=ts_map.get_original_time(segment.end),
)
yield segment
def get_ctranslate2_storage(segment: np.ndarray) -> ctranslate2.StorageView:
segment = np.ascontiguousarray(segment)
segment = ctranslate2.StorageView.from_array(segment)
return segment
def get_compression_ratio(text: str) -> float:
text_bytes = text.encode("utf-8")
return len(text_bytes) / len(zlib.compress(text_bytes))
def get_suppressed_tokens(tokenizer, suppress_tokens):
if not suppress_tokens or -1 in suppress_tokens:
return suppress_tokens
suppress_tokens = list(suppress_tokens)
# Ensure the following special tokens are suppressed when the user does
# not use the default set (-1).
suppress_tokens.extend(
[
tokenizer.transcribe,
tokenizer.translate,
tokenizer.sot,
tokenizer.sot_prev,
tokenizer.sot_lm,
]
)
return sorted(set(suppress_tokens))
def merge_punctuations(alignment: List[dict], prepended: str, appended: str):
# merge prepended punctuations
i = len(alignment) - 2
j = len(alignment) - 1
while i >= 0:
previous = alignment[i]
following = alignment[j]
if previous["word"].startswith(" ") and previous["word"].strip() in prepended:
# prepend it to the following word
following["word"] = previous["word"] + following["word"]
following["tokens"] = previous["tokens"] + following["tokens"]
previous["word"] = ""
previous["tokens"] = []
else:
j = i
i -= 1
# merge appended punctuations
i = 0
j = 1
while j < len(alignment):
previous = alignment[i]
following = alignment[j]
if not previous["word"].endswith(" ") and following["word"] in appended:
# append it to the previous word
previous["word"] = previous["word"] + following["word"]
previous["tokens"] = previous["tokens"] + following["tokens"]
following["word"] = ""
following["tokens"] = []
else:
i = j
j += 1