Gemini 3.8 TTS pricing gives creators a short, cheap window — and then it doubles. Through December 31, 2026, Google lists Gemini 3.8 Flash TTS at $0.50 per million text tokens in and $9 per million audio tokens out, which works out to about $0.81 of audio output per hour of speech. From January 1, 2027, both rates double.
The cheaper Flash-Lite TTS model costs about $0.54 per hour of audio output through December 31, then doubles too. If you make voiceovers or app narration, the next eight weeks are the cheapest time to build your back catalogue.
What Gemini 3.8 TTS actually does
Gemini 3.8 TTS turns a script into speech, and it goes beyond picking a voice from a list. Two features matter for creators:
- Voice Design. Describe a fictional voice in plain words — an older, warm astronomer with a particular accent, for example — and get a new voice identity with a preview.
- Voice Replication. Clone a specific person’s voice, but only with a 10–30 second target recording plus a separate spoken consent recording from the same adult speaker.
Delivery can be directed line by line. A script is treated as a verbatim transcript, while structured metadata carries sustained style, a speaker label, and short acting events such as a laugh, a sigh, or a short pause. Two configured speakers can alternate in one request, which makes dialogue scenes practical without editing two tracks together.
Google says every Gemini Audio clip carries a SynthID watermark. That matters if you sell narration: disclose synthetic audio to clients and platforms, and keep the watermark intact rather than stripping it.
The evidence: prices, languages, and deadlines
| Detail | What the source says |
|---|---|
| Flash TTS through Dec 31 | $0.50 per million text tokens in, $9 per million audio tokens out |
| Flash TTS from Jan 1, 2027 | $1 and $18 — both rates double |
| Flash-Lite TTS through Dec 31 | $0.50 in, $6 out; about $0.54 per hour of audio output |
| Flash-Lite TTS from Jan 1, 2027 | $1 and $12 |
| Audio token rate | 25 audio tokens per second of output, per Google’s pricing page |
| Languages | 130+ for Flash TTS, 101 for Flash-Lite, per launch coverage |
| Prebuilt voices | 30, each tagged with a style such as Bright, Upbeat, or Warm |
| Free try | Free in Google AI Studio unless you link a paid API key |
| Old models retire | Gemini 2.5 and 3.1 TTS models can be switched off from November 17, 2026 |
Token billing means you pay for slices of text going in and slices of audio coming out, not per character of script. That is why a per-hour figure is more useful than a per-word figure for budgeting narration.
Produce a voiceover batch before January 1
1. Pick the model — Flash for expressive narration, Flash-Lite for bulk audio.
2. Design one voice per show — Consistency beats novelty for a series.
3. Script with pauses — Mark laughs, sighs, and short pauses in the metadata.
4. Render the backlog — Queue evergreen episodes now, at 2026 rates.
5. Re-price in January — Re-run the same maths at $1 and $18 before quoting new work.
How to profit from it
Voice work productizes fast, because the client hears the result in seconds:
- Sell a voiceover package, not minutes. Offer ten short-form videos or one audiobook chapter per week, with one consistent designed voice.
- Price from the per-hour maths. At about $0.81 per hour of Flash output, your margin is your time, direction, and editing — not the model. Quote that honestly.
- Offer a two-voice dialogue add-on. Podcasts, ads, and explainer scenes with two speakers are harder to fake and easier to charge for.
- Do not sell cloned voices without consent. Replication needs the owner’s recording and spoken consent. A client who cannot supply both is not a client for that job.
- Lock December delivery dates. Anything rendered after December 31 costs double at the API. Put that deadline in the contract.
What to watch out for
Three cautions before you switch a whole business to this tool. First, the free AI Studio tier may use your text and audio to improve Google’s products — use a paid key for client work. Second, Google’s quality claims are vendor claims, not independent verification. Third, do not confuse Gemini TTS token billing with Google Cloud Text-to-Speech character billing: Chirp 3 HD is listed at $30 per million characters, a different meter.
Sources
Pricing and features as reported from Google’s Gemini API pricing page and launch coverage by Ground Truth, Cartesia, and AI Tool Analysis. Verify the current rate on Google’s pricing page before quoting a client, because the January 1 doubling is already published.
Frequently Asked Questions
How much does Gemini 3.8 Flash TTS cost per hour?
About $0.81 of audio output per hour through December 31, 2026, at $9 per million audio tokens. Input text is extra.
What happens on January 1, 2027?
Flash TTS rates double to $1 per million text tokens and $18 per million audio tokens; Flash-Lite doubles to $1 and $12.
Is Gemini 3.8 TTS free?
It is free in Google AI Studio unless you link a paid API key. The API is token-billed.
Can I clone a real person’s voice?
Only with a 10–30 second target recording plus a separate spoken consent recording from the same adult speaker.
How many languages does it support?
Flash TTS supports 130+ languages and Flash-Lite supports 101, per launch coverage.
Does the audio carry a watermark?
Google says every Gemini Audio clip carries a SynthID watermark, and replicated voices receive C2PA credentials.
What retires on November 17, 2026?
The older Gemini 2.5 and 3.1 TTS models can be switched off from that date, so migrate scripts to the 3.8 models.
Keep Reading
Want more tools that actually pay for themselves? Browse the AI Tools section for the latest launches, pricing breakdowns, and profit playbooks — and check the official source linked above before you buy or deploy anything.