Text to voice generate 13 minutes noise sound for a small text

[qernelwiz]

I just saw that it ha sbeen previously reported:

High Costs Due to Silence or Noisy Segments in gpt-4o-audio-preview Outputs

I wanted to flag an issue we encountered while using the gpt-4o-audio-preview model, which led to unexpected costs and might be worth discussing for those using similar setups. Here’s the situation: We were running the audio-preview model with the temperature parameter set to 0, which is our default for text models. However, this caused the model to emit continuous noise, rather than silence or a logical halt when there was no audio content. This behavior wasn’t explicitly mentioned in the doc…

Realtime API temperature lower than 0.6 bug

Hi, Attached below is my code reference for the Twilio Realtime API (99% identical to the official Twilio example published last month). Unfortunately, any temperature lower than 0.6 results in a “white noise” effect that makes the usability impossible. The white noise “replaces” the bot responses. Has anyone experienced this or knows how to work around it? import os import json import base64 import asyncio import websockets from fastapi import FastAPI, WebSocket, Request from fastapi.respon…