Gemini 3.8 Flash TTS adds Voice Remixing for reusable character voices

Google released Voice Remixing for Gemini 3.8 Flash TTS on November 4. A text prompt can adjust the timbre, pitch, pace or accent of a preset or previously designed voice. The result can be saved as a separate profile. For a game team, that gives NPC casting and later dialogue pickups a common starting point. Google’s release notes

Audition a voice and save the take

In Google AI Studio, pick a starting voice and open Remix this voice. Google’s demo starts with Puck and asks for a warm British accent, slightly lower pitch and unhurried delivery. It shows three takes to audition before saving one. The images are from Google’s demo, which uses simulated screens. Google for Developers

For a branching quest, a team can compare saved profiles using the same reference line, then reuse the chosen voice on alternate routes or for later recordings. The demo shows the chosen voice in the dialogue editor while Puck remains available in the voice library. Google’s release demo

Keep the voice ID for later lines

The Gemini API creates the remixed profile through its Voices endpoint with Gemini 3.8 Flash TTS. It accepts a prebuilt or Extended Voice Library voice, or a stored designed or remixed voice from the same project. The new voice_… ID can synthesize speech with either Gemini 3.8 Flash TTS or Flash-Lite TTS. Voice Remixing guide

The feature requires a paid Gemini API project on Tier 1 or above. Each project can store up to 200 voices across design, remixing and replication. A stored voice expires one year after its last use. Synthesizing with it or using it as the base for another remix restarts that period. For longer productions, keep the chosen ID with the character’s dialogue files and budget speech synthesis at the applicable rate. Voice Remixing guide · API pricing

For related production workflows, see Audio, Music & Voice and Narrative, Dialogue & NPCs.