gemini-3.1-flash-tts-preview replacement: 3.8 Flash-Lite TTS

Google's changelog says gemini-3.8-flash-lite-tts is built to replace gemini-3.1-flash-tts-preview for high-throughput production and real-time voice agent cascades. If you call Sume TTS instead, nothing in your request names a Gemini model: TTS 1.0 has no engine picker, so a Gemini model swap does not change your Sume calls.

Google's facts are from its changelog (Sep 22, 2026) and speech-generation guide; Sume's are from the API reference. Read 2026-10-01.

What does the Gemini change involve?

The Sep 22 changelog entry announces gemini-3.8-flash-tts and gemini-3.8-flash-lite-tts as generally available, plus a Voices endpoint at /v1beta/voices. The guide says to pass the verbatim transcript in input, attach turn-level styling with a speech_metadata annotation, and that unary requests return WAV (audio/wav) with a RIFF header by default. Check the guide itself before you change a production call.

How do I pick an explicit model on Sume?

Use the TTS Router. Its model field is documented as a required TTS Router catalog model id (pass-through). List what is available with GET /v1/tts-router/models, and read the catalog rather than assuming a Gemini id is present.

What stays the same on Sume?

Voices: a voice.id must be a TTS voice UUID or a Voices library id (voi_ plus 32 hex), or you send an avatar reference. Authentication stays a Sume API key; no provider credentials are sent.

What should I do next?

If you call Gemini directly, follow Google's own migration notes. If you call Sume, re-test a sample after any catalog change. Related: which Gemini TTS engine Sume routes.