Gemini 3.8 Flash TTS: Google's Generative Voice Studio

The short version

The new pair reframes text-to-speech as a generative voice studio rather than a fixed preset picker^[1]. Instead of choosing from a short list of voices, you write a natural-language prompt describing a role, accent, and voice characteristics, and the model creates a bespoke voice from scratch^[1]. Google's examples range from a high-energy DJ voice from Melbourne to a fire-breathing Japanese dragon^[1].

Two models, two jobs

• Gemini 3.8 Flash TTS is the flagship, built for "deep creative direction and character design." It's aimed at gaming, immersive audiobooks, podcasts, and interactive media^[1].
• Gemini 3.8 Flash-Lite TTS is the volume sibling, optimized for high-volume dubbing, audio content creation, and expressive voice agents at lower cost^[1].

They join Google's fast-growing Gemini Audio family, which already includes 3.5 Live Translate, 3.5 Transcribe, 3.8 Live, and 3.8 Live Extended Thinking^[1].

Generative voice design

The headline feature is generative voice design: customize role, accent, and voice characteristics across more than 100 languages and dialects using natural-language prompting^[1]. Google also provides an expansive library of 2,000+ production-ready voices with broad language coverage, including regional varieties like Mexican Spanish, Quebec French, and Scots English^[1].

Voice replication recreates a consistent vocal profile from a 30-second audio sample of your own voice or one you have the rights to use, backed by consent verification, SynthID watermarking, and C2PA credentials^[1]. Google notes that voice remixing — picking a library voice and fine-tuning timbre, pitch, pace, and accent — is "coming soon"^[1].

Direct the performance, line by line

Both models give fine-grained control over delivery. You can write your own stage directions or let Gemini steer with script cues — from a calm customer-service agent to a whispered suspense scene^[1]. Other capabilities include^[1]:

• Long-form generation — maintain voice quality and character timbre across hours of audio with minimal speaker drift, ideal for podcasts and audiobooks.
• Native two-speaker scene staging — direct multi-turn conversations from a single script while keeping both voices distinctly separated.
• Scripted vocal bursts and backchanneling — add non-verbal cues like <laughs> or <sigh> and interjections like |mhm| for comedic timing and reaction beats.

Benchmarks

Google reports Gemini 3.8 Flash TTS secured the #1 spot on Hume AI's Voice Design Benchmark (71.4) and led in accent modeling (60.8), with Flash TTS and Flash-Lite TTS taking the #1 and #2 spots respectively on Hume AI's Overall Quality Index^[1]. In blind human-preference evaluations on Voice Arena, both models placed near the top across key languages including Japanese, Brazilian Portuguese, Vietnamese, and Hindi^[1].

Pricing and availability

Both models are rolling out starting September 23, 2026 through the Gemini API and Google AI Studio, with enterprise availability coming soon via Gemini Enterprise^[1]. Flash TTS is also available in Gemini Notebook, and Flash-Lite TTS in Google Vids^[1].

Gemini API pricing for Gemini 3.8 Flash TTS, September 2026; a free tier is listed and batch runs at half price^[3].

Built with consent and watermarking

For voice replication, Google requires a verbal consent recording from the voice owner that matches the reference speaker before a voice can be created^[1]. Every audio clip generated by Gemini Audio models is watermarked with SynthID — an imperceptible watermark woven into the output so AI-generated speech remains detectable^[1]. Full details are in the model card^[4].

Bottom line

Verdict: For podcasters, game developers, audiobook producers, and anyone building voice agents, Gemini 3.8 Flash TTS is a meaningful step up from preset-picker TTS — the ability to design and direct a consistent voice is genuinely useful, and the multilingual coverage is broad. The main caveat is that voice replication, while impressive, carries consent and watermarking guardrails that some workflows may need to plan around.

Frequently asked questions

When did Gemini 3.8 Flash TTS release?

Google announced both models on September 23, 2026^[1].

How many languages does it support?

More than 100 languages and dialects, with 2,000+ production-ready voices including regional varieties^[1].

Can it clone a voice?

Yes — voice replication works from a 30-second audio sample, with consent verification, SynthID watermarking, and C2PA credentials^[1].

How much does Gemini 3.8 Flash TTS cost?

Introductory pricing through December 31, 2026 is $0.50 per million text input tokens and $9.00 per million audio output tokens, rising to $1.00 / $18.00 from January 1, 2027^[3].