functions/speech_to_text.py

from faster_whisper import WhisperModel
import base64, tempfile, os
import torch

model = WhisperModel(
    "base",
    device="cuda" if torch.cuda.is_available() else "cpu",
    compute_type="float16" if torch.cuda.is_available() else "int8",
)


def speechToText(audioBase64: str, sourceLang: str) -> dict:
    tempAudioPath = None
    try:
        audioBytes = base64.b64decode(audioBase64)

        with tempfile.NamedTemporaryFile(delete=False, suffix=".m4a") as tempFile:
            tempFile.write(audioBytes)
            tempAudioPath = tempFile.name
        
        segments, info = model.transcribe(tempAudioPath, language=sourceLang)
        text = " ".join(segment.text for segment in segments)
        return {'text': text, 'language': info.language, 'duration': info.duration}
    finally:
        if tempAudioPath and os.path.exists(tempAudioPath):
            os.remove(tempAudioPath)