[yoyoismee]

Summary

OpenAI is shutting down tts-1 and tts-1-hd, which the OpenAI TTS page recommends, and whisper-1, which the OpenAI STT page still lists. Neither page mentions it yet. This adds a short caution to each page, with the dates from OpenAI's deprecations page:

• TTS: tts-1, tts-1-hd and gpt-4o-mini-tts are removed on 2027-01-06. OpenAI's replacement, gpt-realtime-2.1-mini, isn't supported on /v1/audio/speech, so changing the model in the OpenAI engine won't help. The caution points to the OpenAI Realtime engine, which the page already documents and which defaults to that model.
• STT: whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe and gpt-4o-transcribe-diarize are removed on 2027-02-26. The caution says to switch to gpt-transcribe, which the page already recommends.

Two related fixes on the same pages:

• The STT page said Open WebUI uses whisper-1 when AUDIO_STT_MODEL is unset, and its env table listed gpt-transcribe as the default. The code does neither: the default is an empty string, and _transcribe_openai sends it as is. The page now says to set the model explicitly, and the table says empty, like the environment variable reference.
• Each troubleshooting list gets one step for people who point the OpenAI engine at another OpenAI-compatible service: use that service's model and voice IDs, because the defaults (tts-1, alloy) are OpenAI's.

Related issue or discussion

None.

Checklist

• I have reviewed the relevant documentation and matched the existing style.
• This PR meets Open WebUI's contribution standards: it is accurate, relevant to users, narrowly scoped, maintainable, and not promotional content, advertising, lead generation, SEO placement, or a request to list a product, service, provider, integration, gateway, tool, or company primarily for visibility.
• I understand that PRs that do not meet these standards may be closed without review and will not be merged. Repeated, low-quality, off-topic, promotional, or intentionally misleading submissions may result in the contributor being blocked from future participation in Open WebUI repositories.

Notes for reviewers

• Sources: OpenAI's deprecations page (the entries dated 2026-10-01 and 2026-08-26) and its text-to-speech guide, which says "gpt-realtime-2.1-mini isn't a supported model for /v1/audio/speech".
• Checked against open-webui main (the same on dev): AUDIO_TTS_MODEL and AUDIO_TTS_VOICE default to tts-1 and alloy, and selecting the OpenAI engine in Audio.svelte fills in the same pair. AUDIO_STT_MODEL defaults to ''. The OpenAI engine posts to {base}/audio/speech.
• I left the other tts-1 mentions alone: the quality, rate-limit and cost notes, and the quick setup on the audio troubleshooting page, stay accurate until 2027-01-06, and tts-1/alloy are still the backend defaults. Happy to update them in this PR if you'd rather do it in one pass.
• No headings were added or renamed. Both pages compile as MDX with GFM and directives, and the new #using-the-openai-realtime-engine link resolves. I didn't run the full site build locally.