openai_audio_streaming.py

import base64
import os
from openai import OpenAI
import pyaudio
import wave
import weave

os.environ["WANDB_API_KEY"] = "" # Set your wandb API key here to track audio
client = OpenAI(api_key=os.environ.get("OPENAI_API_KEY"))

weave.init('openai-audio-chat')
@weave.op
def get_audio_stream():
    # Initialize PyAudio
    p = pyaudio.PyAudio()
    
    # Open an audio stream
    stream = p.open(format=pyaudio.paInt16,
                    channels=1,
                    rate=22050,
                    output=True,
                    frames_per_buffer=4096)

    # Open a wave file for writing
    with wave.open("output.wav", "wb") as wav_file:
        wav_file.setnchannels(1)  # Mono
        wav_file.setsampwidth(2)  # 16-bit
        wav_file.setframerate(22050)  # Sample rate (adjust if needed)

        completion = client.chat.completions.create(
            model="gpt-4o-audio-preview",
            modalities=["text", "audio"],
            audio={"voice": "fable", "format": "pcm16"},
            stream=True,
            messages=[
                {"role": "system", "content": "You're the fastest counter in the world"},
                {"role": "user", "content": "Count to 13 super super slow, announciate each number with a dramatic flair, chnging up accents as you go along. British, french, german, spanish, etc."}
            ]
        )

        for chunk in completion:
            if hasattr(chunk, 'choices') and chunk.choices is not None and len(chunk.choices) > 0:
                if hasattr(chunk.choices[0].delta, 'audio') and chunk.choices[0].delta.audio.get('data') is not None:
                    audio_data = base64.b64decode(chunk.choices[0].delta.audio.get('data'))
                    
                    # Write to audio stream
                    stream.write(audio_data)
                    
                    # Write to wave file
                    wav_file.writeframes(audio_data)

    # Close and terminate PyAudio
    stream.stop_stream()
    stream.close()
    p.terminate()

    return wave.open("output.wav", "rb")

# Call the function
get_audio_stream()

requirements.txt

openai
weave
pyaudio