{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreictwkprllxushad7qr5n5uf4rcl2h5r6wqwdwvwlvt4rff5yu4iyy",
    "uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mhihpagozof2"
  },
  "path": "/t/gpt-realtime-1-5-major-regression-in-voice-expressiveness-and-accent-quality/1377222#post_3",
  "publishedAt": "2026-03-20T10:30:10.000Z",
  "site": "https://community.openai.com",
  "textContent": "Hello, yes I have noticed the same thing. In my scenario the 1.5 model changes the pitch and speed of the agents voice for seemingly no reason. This doesn’t happen on the standard realtime model, and I have not been able to address these issues through prompting. I agree the model is a clear downgrade at the present time, I am sticking to gpt-realtime until the 1.5 issues are resolved. There doesn’t appear to have been any acknowledgement from OpenAI however?",
  "title": "GPT-Realtime 1.5 — Major regression in voice expressiveness and accent quality"
}