API
Google is retiring five preview audio models from the Gemini API on 17 November 2026 — the text-to-speech models gemini-3.1-flash-tts-preview, gemini-2.5-flash-preview-tts and gemini-2.5-pro-preview-tts, and the Live API models gemini-3.1-flash-live-preview and gemini-2.5-flash-native-audio-preview-12-2025 — directing developers to the gemini-3.8 Live and TTS successors
departs 17 Nov 2026
Quick answers
- Is it true that Google shuts down five Gemini API preview TTS and Live audio models on 17 November 2026?
- Yes. Google has officially announced it, effective 17 Nov 2026. Google's official Gemini API deprecations page lists five preview audio models with a shutdown date of **17 November 2026**.
- When does it take effect?
- 17 Nov 2026. Verified against the vendor's own announcement.
- What should I use instead?
- Repoint each retiring model to its Google-named `gemini-3.8` successor before 17 November 2026.
What this means for you
Google's official Gemini API deprecations page lists five preview audio models with a shutdown date of **17 November 2026**. Three are text-to-speech models — `gemini-3.1-flash-tts-preview`, `gemini-2.5-flash-preview-tts` and `gemini-2.5-pro-preview-tts` — and two are Live API (real-time audio) models — `gemini-3.1-flash-live-preview` and `gemini-2.5-flash-native-audio-preview-12-2025`. These are the preview-labelled models developers have been using to build voice output and live voice conversation on the Gemini API, and all five stop being served on the same day. After the shutdown, a request that names one of these model IDs fails rather than falling back to anything, so any app that generates speech or runs a live audio session through these models breaks at the API call unless it has been repointed. The "preview" label was always the warning that this day would come — preview models carry no stability guarantee — but because voice and real-time audio are often wired deep into an app's interaction layer, the migration is rarely a one-line change.
Where to go
Repoint each retiring model to its Google-named `gemini-3.8` successor before 17 November 2026. For the TTS models (`gemini-3.1-flash-tts-preview`, `gemini-2.5-flash-preview-tts`, `gemini-2.5-pro-preview-tts`), Google names **`gemini-3.8-flash-tts` or `gemini-3.8-flash-lite-tts`**. For the Live API models (`gemini-3.1-flash-live-preview`, `gemini-2.5-flash-native-audio-preview-12-2025`), the named replacement is **`gemini-3.8-live`**. These are different model generations, not drop-in swaps: voice selection, audio format and streaming behaviour, latency, and pricing can all move between versions, so test the successor against your actual prompts and audio pipeline rather than assuming parity — especially for Live API sessions, where session setup and the real-time event stream are easy to break silently. Because these are preview models, confirm the current successor IDs on the official deprecations page before you cut over; Google adjusts this table as models graduate out of preview.
- gemini-3.8-flash-tts or gemini-3.8-flash-lite-tts — Google's named replacements for the three retiring Gemini TTS preview models
- gemini-3.8-live — Google's named replacement for both retiring Gemini Live API preview models
Google’s official Gemini API deprecations page lists five preview audio models with a shutdown date of 17 November 2026. All five stop being served that day, and a request naming one of their IDs will fail rather than fall back to anything.
The five models retiring on 17 November 2026
Text-to-speech:
gemini-3.1-flash-tts-previewgemini-2.5-flash-preview-ttsgemini-2.5-pro-preview-tts
Live API (real-time audio):
gemini-3.1-flash-live-previewgemini-2.5-flash-native-audio-preview-12-2025
What this means for you
These are the preview models developers have used to build voice output (TTS) and live voice conversation (the Live API) on top of Gemini. Because they are preview-labelled, they never carried a stability guarantee — but voice and real-time audio tend to sit deep in an app’s interaction layer, so pulling them out is rarely a one-line change. After 17 November 2026, any call that still names one of these IDs breaks at the API.
Migration
Repoint to the Google-named gemini-3.8 successors before the date:
- TTS →
gemini-3.8-flash-ttsorgemini-3.8-flash-lite-tts - Live API →
gemini-3.8-live
Treat this as a generation change, not a drop-in swap: voice selection, audio format and streaming behaviour, latency and pricing can all differ. Test the successor against your real prompts and audio pipeline — especially for Live API sessions, where session setup and the real-time event stream are easy to break silently. Confirm the current successor IDs on the deprecations page before cutover; Google adjusts this table as preview models graduate.
Entry changelog
- — last reviewed.
- — Entry created from Google's official Gemini API deprecations page (ai.google.dev/gemini-api/docs/deprecations), which lists five models with a shutdown date of 17 November 2026: TTS models gemini-3.1-flash-tts-preview, gemini-2.5-flash-preview-tts and gemini-2.5-pro-preview-tts (replacement gemini-3.8-flash-tts or gemini-3.8-flash-lite-tts), and Live API models gemini-3.1-flash-live-preview and gemini-2.5-flash-native-audio-preview-12-2025 (replacement gemini-3.8-live). Date and model IDs recorded exactly as stated; no date inferred (Constitution §3.2). detect: fingerprints are the five retiring model IDs only — the gemini-3.8 replacements are deliberately NOT fingerprinted (never fingerprint the successor, SOURCING.md). Distinct from the existing gemini-2-5-models-eol entry, which covers the Gemini 2.5 text/image generation, not these audio preview models.
Put this countdown in your README
Building on Google? Drop this badge into a project’s README or docs — it shows the days left and
refreshes on its own every night from this page. 
[](https://lastcall.dev/entries/gemini-tts-live-audio-preview-models-eol)Spot an error or a changed date? Tell us and it’s corrected the same night.