Transcribe audio using OpenAI

transcribe.py

import argparse
import os
from openai import OpenAI
from pydub import AudioSegment


def chunk_audio_by_size(audio, max_size_bytes=25000000, target_duration_ms=60000):
    """
    Splits an audio file into smaller chunks based on a maximum size limit.
    Args:
        audio (AudioSegment): Audio file as a pydub AudioSegment.
        max_size_bytes (int): Maximum size of each chunk in bytes.
        target_duration_ms (int): Target duration of chunks in milliseconds.
    Returns:
        list of AudioSegment: List of audio chunks.
    """
    chunks = []
    start = 0

    while start < len(audio):
        end = start + target_duration_ms
        chunk = audio[start:end]
        # Ensure chunk size is within the limit
        while len(chunk.raw_data) > max_size_bytes and target_duration_ms > 1000:
            target_duration_ms //= 2  # Halve the duration to reduce size
            chunk = audio[start:start + target_duration_ms]
        if len(chunk.raw_data) > max_size_bytes:
            raise ValueError("Cannot create a chunk under the size limit.")
        chunks.append(chunk)
        start += target_duration_ms

    return chunks


def transcribe_audio_chunk(client, audio_chunk, temp_file_path):
    """
    Transcribes a single audio chunk using OpenAI API.
    Args:
        client (OpenAI): Initialized OpenAI client.
        audio_chunk (AudioSegment): Audio chunk to transcribe.
        temp_file_path (str): Temporary file path to save the audio chunk.
    Returns:
        str: Transcription text for the chunk.
    """
    audio_chunk.export(temp_file_path, format="wav")
    with open(temp_file_path, "rb") as audio_file:
        response = client.audio.transcriptions.create(
            model="whisper-1",
            file=audio_file
        )
    if hasattr(response, 'text'):
        return response.text
    else:
        raise ValueError("Transcription text attribute not found in the response")


def transcribe_audio(audio_path):
    # Read the OpenAI API key from an environment variable
    openai_api_key = os.environ.get('OPENAI_API_KEY')
    if not openai_api_key:
        raise ValueError("OPENAI_API_KEY environment variable not set")

    # Initialize the OpenAI client with the API key
    client = OpenAI(api_key=openai_api_key)

    # Load and split the audio file
    audio = AudioSegment.from_file(audio_path)
    audio_chunks = chunk_audio_by_size(audio)

    transcription_parts = []
    temp_file_path = "temp_chunk.wav"

    for i, chunk in enumerate(audio_chunks):
        print(f"Transcribing chunk {i + 1}/{len(audio_chunks)}...")
        try:
            transcription_part = transcribe_audio_chunk(client, chunk, temp_file_path)
            transcription_parts.append(transcription_part)
        except Exception as e:
            print(f"Error transcribing chunk {i + 1}: {e}")
            transcription_parts.append("[Error in transcription]")

    # Combine all transcription parts
    transcription_text = "\n".join(transcription_parts)
    os.remove(temp_file_path)  # Clean up the temporary file

    return transcription_text


def main():
    # Parse command line arguments
    parser = argparse.ArgumentParser(description='Transcribe audio file using OpenAI API.')
    parser.add_argument('--audio-file', type=str, required=True, help='Path to the audio file')
    args = parser.parse_args()

    # Transcribe the audio
    transcription_text = transcribe_audio(args.audio_file)

    # Save the transcription to a text file
    output_file = args.audio_file + '.txt'
    with open(output_file, 'w') as file:
        file.write(transcription_text)

    print(f"Transcription saved to {output_file}")


if __name__ == '__main__':
    main()