Welcome to Whisper Live documentation!

class whisper_live.server.ServeClient(websocket, task='transcribe', device=None, multilingual=False, language=None, client_uid=None)
Attributes:

RATE (int): The audio sampling rate (constant) set to 16000. SERVER_READY (str): A constant message indicating that the server is ready. DISCONNECT (str): A constant message indicating that the client should disconnect. client_uid (str): A unique identifier for the client. data (bytes): Accumulated audio data. frames (bytes): Accumulated audio frames. language (str): The language for transcription. task (str): The task type, e.g., “transcribe.” transcriber (WhisperModel): The Whisper model for speech-to-text. timestamp_offset (float): The offset in audio timestamps. frames_np (numpy.ndarray): NumPy array to store audio frames. frames_offset (float): The offset in audio frames. text (list): List of transcribed text segments. current_out (str): The current incomplete transcription. prev_out (str): The previous incomplete transcription. t_start (float): Timestamp for the start of transcription. exit (bool): A flag to exit the transcription thread. same_output_threshold (int): Threshold for consecutive same output segments. show_prev_out_thresh (int): Threshold for showing previous output segments. add_pause_thresh (int): Threshold for adding a pause (blank) segment. transcript (list): List of transcribed segments. send_last_n_segments (int): Number of last segments to send to the client. wrapper (textwrap.TextWrapper): Text wrapper for formatting text. pick_previous_segments (int): Number of previous segments to include in the output. websocket: The WebSocket connection for the client.

add_frames(frame_np)

Add audio frames to the ongoing audio stream buffer.

This method is responsible for maintaining the audio stream buffer, allowing the continuous addition of audio frames as they are received. It also ensures that the buffer does not exceed a specified size to prevent excessive memory usage.

If the buffer size exceeds a threshold (45 seconds of audio data), it discards the oldest 30 seconds of audio data to maintain a reasonable buffer size. If the buffer is empty, it initializes it with the provided audio frame. The audio stream buffer is used for real-time processing of audio data for transcription.

Args:

frame_np (numpy.ndarray): The audio frame data as a NumPy array.

cleanup()

Perform cleanup tasks before exiting the transcription service.

This method performs necessary cleanup tasks, including stopping the transcription thread, marking the exit flag to indicate the transcription thread should exit gracefully, and destroying resources associated with the transcription process.

disconnect()

Notify the client of disconnection and send a disconnect message.

This method sends a disconnect message to the client via the WebSocket connection to notify them that the transcription service is disconnecting gracefully.

fill_output(output)

Format the current incomplete transcription output by combining it with previous complete segments. The resulting transcription is wrapped into two lines, each containing a maximum of 50 characters.

It ensures that the combined transcription fits within two lines, with a maximum of 50 characters per line. Segments are concatenated in the order they exist in the list of previous segments, with the most recent complete segment first and older segments prepended as needed to maintain the character limit. If a 3-second pause is detected in the previous segments, any text preceding it is discarded to ensure the transcription starts with the most recent complete content. The resulting transcription is returned as a single string.

Args:

output(str): The current incomplete transcription segment.

Returns:

str: A formatted transcription wrapped in two lines.

speech_to_text()

Process an audio stream in an infinite loop, continuously transcribing the speech.

This method continuously receives audio frames, performs real-time transcription, and sends transcribed segments to the client via a WebSocket connection.

If the client’s language is not detected, it waits for 30 seconds of audio input to make a language prediction. It utilizes the Whisper ASR model to transcribe the audio, continuously processing and streaming results. Segments are sent to the client in real-time, and a history of segments is maintained to provide context.Pauses in speech (no output from Whisper) are handled by showing the previous output for a set duration. A blank segment is added if there is no speech for a specified duration to indicate a pause.

Raises:

Exception: If there is an issue with audio processing or WebSocket communication.

update_segments(segments, duration)

Processes the segments from whisper. Appends all the segments to the list except for the last segment assuming that it is incomplete.

Updates the ongoing transcript with transcribed segments, including their start and end times. Complete segments are appended to the transcript in chronological order. Incomplete segments (assumed to be the last one) are processed to identify repeated content. If the same incomplete segment is seen multiple times, it updates the offset and appends the segment to the transcript. A threshold is used to detect repeated content and ensure it is only included once in the transcript. The timestamp offset is updated based on the duration of processed segments. The method returns the last processed segment, allowing it to be sent to the client for real-time updates.

Args:

segments(dict) : dictionary of segments as returned by whisper duration(float): duration of the current chunk

Returns:
dict or None: The last processed segment with its start time, end time, and transcribed text.

Returns None if there are no valid segments to process.

class whisper_live.server.TranscriptionServer

Represents a transcription server that handles incoming audio from clients.

Attributes:

RATE (int): The audio sampling rate (constant) set to 16000. vad_model (torch.Module): The voice activity detection model. vad_threshold (float): The voice activity detection threshold. clients (dict): A dictionary to store connected clients. websockets (dict): A dictionary to store WebSocket connections. clients_start_time (dict): A dictionary to track client start times. max_clients (int): Maximum allowed connected clients. max_connection_time (int): Maximum allowed connection time in seconds.

get_wait_time()

Calculate and return the estimated wait time for clients.

Returns:

float: The estimated wait time in minutes.

recv_audio(websocket)

Receive audio chunks from a client in an infinite loop.

Continuously receives audio frames from a connected client over a WebSocket connection. It processes the audio frames using a voice activity detection (VAD) model to determine if they contain speech or not. If the audio frame contains speech, it is added to the client’s audio data for ASR. If the maximum number of clients is reached, the method sends a “WAIT” status to the client, indicating that they should wait until a slot is available. If a client’s connection exceeds the maximum allowed time, it will be disconnected, and the client’s resources will be cleaned up.

Args:

websocket (WebSocket): The WebSocket connection for the client.

Raises:

Exception: If there is an error during the audio frame processing.

run(host, port=9090)

Run the transcription server.

Args:

host (str): The host address to bind the server. port (int): The port number to bind the server.

class whisper_live.client.TranscriptionClient(host, port, is_multilingual=False, lang=None, translate=False)

Client for handling audio transcription tasks via a 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.

Args:

host (str): The hostname or IP address of the server. port (int): The port number to connect to on the server. is_multilingual (bool, optional): Indicates whether the transcription should support multiple languages (default is False). lang (str, optional): The primary language for transcription (used if is_multilingual is False). Default is None, which defaults to English (‘en’). translate (bool, optional): Indicates whether translation tasks are required (default is False).

Attributes:

client (Client): An instance of the underlying Client class responsible for handling the WebSocket connection.

Example:

To create a TranscriptionClient and start transcription on microphone audio: `python transcription_client = TranscriptionClient(host="localhost", port=9090, is_multilingual=True) transcription_client() `

whisper_live.client.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

Indices and tables