Gemini Voice Remixing is a feature Google added to Gemini 3.8 Flash TTS on November 4, 2026: pick an existing voice, say in one sentence how it should change — older, calmer, a different accent — and you get a new voice you can save and reuse. The entry points are Google AI Studio and the Gemini API, and the remixed voice works in both Flash TTS and Flash-Lite TTS.
Remix a voice in four steps
Step one: open Google AI Studio and go to the TTS interface. Step two: pick a starting point — any of the 2,000+ preset voices, or a voice you designed yourself earlier. Step three: click Remix this voice and write one sentence describing the change, for example, "a gentler tone, a slower pace"; changing one or two traits at a time works best, and changing more tends to go off track. Step four: listen to the three generated versions side by side, pick one, and save it to get an ID starting with voice_, which you then use directly when synthesizing speech. Via the API, call POST /v1beta/voices to get the same reusable voice.
Limits: free tier, count, and expiry
Remixing voices through the API requires a paid Tier 1 or higher project; the free tier cannot do it. Each project can store at most 200 custom voices, kept for one year from last use, after which unused ones are cleaned up. Do not confuse this with copying a real person's voice: that is a separate feature, Voice Replication, which needs a 30-second sample plus a verbal consent recording from the person. Remix changes an existing AI voice and does not touch a real identity.
When to use it, and when not to
It fits best for fixed characters and fixed shows: tune the voice once, then reuse it in every episode without describing it again. To change the tone of a single line or shift a mood on the fly, skip Remix and write a style hint directly in the speech parameters — it is faster. Anyone wanting to change five or six traits at once and get it right the first time should go in two or three rounds, moving only one or two points per round.
