2026-11-16 18:26 UTC
DANGMUAAI & Developer Tools, Decoded
BackAI Models

Gemini TTS Preview Models Can Shut Down From Nov 17

Google's deprecation page lists November 17, 2026 as the earliest shutdown date for five Gemini audio preview models. Here are the IDs and their replacements.

DangMua EditorialNov 16, 20263 min read
Gemini TTS Preview Models Can Shut Down From Nov 17

Google's deprecation page lists November 17, 2026 as the earliest shutdown date for five Gemini audio preview models. That deadline is here, and "earliest" is doing a lot of work in that sentence.

The list comes from a Dev.to write-up by Versyka, a vendor that tracks provider shutdown dates — worth noting, since the post also pitches its own product. The model IDs and replacements it published are below.

The five IDs

Model IDMove to
gemini-3.1-flash-tts-previewgemini-3.8-flash-tts or gemini-3.8-flash-lite-tts
gemini-2.5-flash-preview-ttssame
gemini-2.5-pro-preview-ttssame
gemini-3.1-flash-live-previewgemini-3.8-live
gemini-2.5-flash-native-audio-preview-12-2025gemini-3.8-live

Why "earliest" changes your planning

A fixed date gives you a deadline you can schedule against. An earliest date does not: per the source, it means the calls can stop on any day from then on, without another announcement. The post also notes that preview models tend to get shorter notice than stable ones, because they were never promised to last in the first place.

Practically, that turns a migration ticket into something you either do this week or discover through an error log.

Audio is the easy one to miss

The source makes a fair argument about why this specific deprecation slips past teams: voice features get built fast — a read-aloud button, a voice mode for a demo, a live agent prototype that quietly became a feature. They are rarely the core of the app, so nobody owns the migration until the first bug report arrives.

What to do now

Start with a search across code and config. The source suggests:

grep -rnE "preview-tts|native-audio-preview|flash-live-preview" .

Then check environment variables and anywhere a model name is stored as data rather than code — config tables, feature flags, tenant settings. A grep over the repo will not find a model ID that lives in a database row.

For the swap itself: move TTS calls to gemini-3.8-flash-tts, or gemini-3.8-flash-lite-tts if cost matters more than quality, and Live calls to gemini-3.8-live.

Two checks that catch what tests miss

Listen to the output. Voices, pacing and pronunciation change between models, and as the source puts it, a test that only checks "audio came back" will not catch a voice your users do not recognise. If your product has a recognisable voice, that is a product change, not just a config change.

Then run one real Live session end to end. Session settings that worked on a preview may not carry over unchanged — and a Live session failing mid-call is a worse failure mode than a TTS request returning an error.

The wider pattern

Preview model IDs are not a stable foundation. If a preview ID is running in production in your app right now, the migration is not the risk — the risk is that nobody currently knows it is there.

More from DangMua