backend/voice/tts.py

import openai

client = openai.AsyncOpenAI()

async def synthesize_speech(text: str, voice: str = "alloy") -> bytes:
    # Use tts-1 for speed (not tts-1-hd)
    tts_resp = await client.audio.speech.create(
        model="tts-1",  # Faster than tts-1-hd
        voice=voice,
        input=text,
        response_format="mp3",  # More efficient than default
        speed=1.1  # Slightly faster speech
    )
    return tts_resp.content