Feature: Transcription Cancellation

[rudymohammadbali]

Overview

I'm excited to share a new feature that I've added to the Whisper transcribe module, allowing users to cancel the transcription process using a customizable cancellation function. This feature can be handy in scenarios where users need to interrupt ongoing transcriptions for various reasons.

Changes Made

I've made the following changes in the whisper.transcribe module:

def transcribe(
    # ... (other parameters)
    cancel_func: any,  # added param 
    # ... (other parameters)
):
    # ... (existing code)

    # added check for the signals to break the loop
    while seek < content_frames:
        signal = cancel_func()
        if signal:
            print("Transcription canceled.")
            return dict(text="", segments=[], language=language)
            break

# ... (other existing code)

Usage Example

import whisper

# Define a global variable to serve as a cancellation flag
CANCEL_FLAG = False

# Function to set the cancellation flag
def cancel_transcription():
    global CANCEL_FLAG
    CANCEL_FLAG = True
    print("Transcription cancellation requested.")
    print("-" * 50)

# Function to check if cancellation is requested
def cancel_requested():
    global CANCEL_FLAG
    print("Cancellation requested status:", CANCEL_FLAG)
    original_value = CANCEL_FLAG  # Save the original value

    if CANCEL_FLAG:
        # Reset the cancellation flag only if it is True
        CANCEL_FLAG = False
        print("Cancellation flag reset to False.")

    print("-" * 50)

    return original_value  # Return the original value

# ... (other code)

# Example of using the new cancellation feature
print("Transcribing....")
model = whisper.load_model("base")
result = model.transcribe("E:\\Python\\WhisperTranscriber\\assets\\audios\\1min.wav", cancel_func=cancel_requested)
print(result["text"])