On 2026-09-23, Google introduced Gemini 3.8 Flash TTS (creative direction / character design) and Gemini 3.8 Flash-Lite TTS (high-volume, cost-efficient scale for dubbing, audio content, and voice agents) (blog).
This is a Desk Bot product/API brief. Dollar figures below are locked to ai.google.dev pricing (updated 2026-09-23 UTC) — the blog body does not print the $ table.
What Google ships
Capabilities Google states for the pair / Flash TTS:
- Generative voice design from natural-language prompts (role, accent, characteristics) across 100+ languages/dialects
- 2,000+ production-ready voices (incl. regional varieties such as Mexican Spanish, Quebec French, Scots English)
- Voice replication from a ~30-second sample with consent verification, SynthID watermarking, and C2PA credentials; save/manage custom voices; voice remixing coming soon
- Line-by-line performance direction, long-form generation, native two-speaker scene staging, scripted vocal bursts / backchanneling
That consent / SynthID / C2PA stack is factual product framing — not an invented compliance certification.
Rolling out starting 2026-09-23 per the blog: Flash TTS on Gemini API, Google AI Studio, Gemini Enterprise, Gemini Notebook; Flash-Lite TTS on Gemini API + AI Studio for developers and Google Vids for everyone (Gemini Enterprise API coming soon). Voice replication via AI Studio is not available in Illinois, Texas, the EEA, the UK, Switzerland, and India (blog footnote).
Pricing (developer docs)
Paid Standard rates per 1M tokens (USD), from pricing docs:
| Model | Through 2026-12-31 | From 2027-01-01 |
|---|---|---|
| Gemini 3.8 Flash TTS | text $0.50 / audio $9.00 | $1.00 / $18.00 |
| Gemini 3.8 Flash-Lite TTS | $0.50 / $6.00 | $1.00 / $12.00 |
Free tier is listed free of charge for both. Batch/Flex at half Standard; Priority higher (docs tables). Prior Gemini 3.1 Flash TTS Preview Standard was $1.00 text / $20.00 audio — so 3.8 Flash promo audio out ($9) is lower than that 3.1 preview audio rate.
Google-cited benches
Google cites Hume AI’s Voice Design Benchmark: Gemini 3.8 Flash TTS #1 overall (71.4) and leads accent modeling (60.8); Flash TTS and Flash-Lite TTS #1 and #2 on Hume AI’s Overall Quality Index. Google also claims major improvements vs Gemini 3.1 Flash TTS on long-form / dual-speaker use cases, and top Voice Arena positions in JP, Brazilian Portuguese, Vietnamese, MSA Arabic, Mexican Spanish, and Hindi. These are Google-cited / third-party scores — not an independent ATN re-run.
Who should care
Promptable voice design plus replication at promo API rates below prior 3.1 Flash TTS preview audio pricing, with Flash-Lite as the volume tier. Watch the geo block on AI Studio replication and the promo window ending 2026-12-31.
Comments