# tts-1-hd

Calls to the model tts-1-hd on OpenAI stop working on January 6, 2027.

On January 6, 2027, OpenAI will shut down the model tts-1-hd. Any requests specifying model tts-1-hd will fail after that date.

## The error OpenAI returns once tts-1-hd is gone

HTTP 404

```json
{"error":{"message":"The model `tts-1-hd` has been deprecated, learn more here: https://platform.openai.com/docs/deprecations","type":"invalid_request_error","param":null,"code":"model_not_found"}}
```

## What to use instead

Migrate to the recommended replacement gpt-realtime-2.1-mini. This transition requires more than updating a model identifier, as gpt-realtime-2.1-mini is available only on the Realtime API at /v1/realtime using WebRTC or WebSocket rather than /v1/audio/speech, and it returns audio as base64 chunks in response.output_audio.delta events instead of mp3 or wav files.

gpt-realtime-2.1-mini is not served on /v1/audio/speech: it runs only on the Realtime API (/v1/realtime, over WebRTC or WebSocket). Audio comes back as base64 chunks in response.output_audio.delta events (for example audio/pcm at 24 kHz), not as an mp3 or wav response body.

- API: OpenAI (api.openai.com)
- Model: tts-1-hd
- Released: November 6, 2023 (models.dev)
- Kind: Model retirement
- Takes effect: January 6, 2027
- Calls after that day: fail
- Calls it hits: `model tts-1-hd`
- Source: [Deprecations page](https://developers.openai.com/api/docs/deprecations)
- Verified: November 6, 2026, against OpenAI's own pages
- Page: https://www.okapi.bot/models/openai/tts-1-hd/

## OpenAI's words

> Shutdown date: Jan 6, 2027; Model / system: tts-1-hd; Recommended replacement: gpt-realtime-2.1-mini

Okapi finds the calls this change hits in your repo and opens the pull request with the fix: https://github.com/apps/okapi-agent