backend/voice/stt.py

import openai
import io

# Use async client for better performance
client = openai.AsyncOpenAI()

async def transcribe_audio(audio_bytes: bytes, file_ext: str = ".m4a") -> str:
    file_obj = io.BytesIO(audio_bytes)
    file_obj.name = "audio" + file_ext
    
    # Add language hint and prompt for faster processing
    transcript_resp = await client.audio.transcriptions.create(
        model="whisper-1",
        file=file_obj,
        response_format="text",
        language="en",  # Hint for faster processing
        prompt="This is a conversation about wellness and mental health."  # Context helps
    )
    return transcript_resp