Changing a voice’s baseline and asking it to deliver one sentence calmly are different production decisions.
Google added voice remixing to Gemini 3.8 TTS on November 4, according to its API release notes. A developer can describe changes to an existing voice and create a separate stored voice for later speech generation. The Flash and Flash-Lite TTS models became generally available in September.
A remix receives its own stored identity
The voice-remixing guide uses POST /v1beta/voices with gemini-3.8-flash-tts. The result has a new voice_... identifier that can be used with either gemini-3.8-flash-tts or gemini-3.8-flash-lite-tts.
The base can be a prebuilt voice, a voice from the extended library, or a supported project-owned prompted voice. A relative prompt can change qualities such as pitch, energy, timbre, or accent. Google also offers an interactive route in AI Studio.
A stored remix is independent of its source. It does not overwrite the original, and deleting the source does not delete the remix. Teams therefore need to track the new identifier and manage its lifecycle explicitly.
Persistent voice and per-turn delivery serve different jobs
Remixing changes the voice baseline. The speech_metadata.style instruction can direct a particular passage’s delivery. For example, a production may reuse one voice identity while requesting a calmer delivery for a later scene.
MustHave has not generated or listened to the remixed output. A useful evaluation would compare the same script across the source voice and remix, then check pronunciation and stability over several passages.
Price the audio for the month it runs
| Standard paid audio output | Through December 31, 2026 | From January 1, 2027 |
|---|---|---|
| Gemini 3.8 Flash TTS | $9 per million audio tokens | $18 per million audio tokens |
| Gemini 3.8 Flash-Lite TTS | $6 per million audio tokens | $12 per million audio tokens |
Google’s pricing page states that audio uses 25 tokens per second. One generated minute therefore uses about 1,500 audio tokens. At current standard output rates, that is about $0.0135 for Flash or $0.009 for Flash-Lite, before text input, retries, or other costs. These are calculated output-only estimates.
Keep voice creation and repeated synthesis separate in a production budget. Store the approved voice ID with the script and model configuration so later changes can be reviewed and, if needed, reverted.
Abdessalam Alaoui
I'm Abdessalam Alaoui, founder of Musthave.AI. I write practical guides to AI models, tools, generative media, and ways real people make money with this stuff. No press-release fluff. 20+ years shipping digital products; you'll also find my work around HowTik, Yumasty, and IncomeHelpful.
