What it is
Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS became generally available on the Gemini API on 22 September 2026, alongside a new Voices endpoint. You can design a voice from a text description, replicate a voice from a 30-second sample plus a spoken consent recording, direct delivery line by line, and add tags such as laughs and sighs. Flash-Lite replaces gemini-3.1-flash-tts-preview.
Flash TTS covers 130+ languages and targets audiobooks, games and podcasts; Flash-Lite covers 100+ and targets dubbing and voice agents. All output carries Google's SynthID watermark, and replicated voices also get C2PA credentials. Voice replication in AI Studio is not offered in Illinois, Texas, the EEA, the UK, Switzerland or India.
Our review
Gemini 3.8 Flash TTS gets close to ElevenLabs for a fraction of the cost, at least until January. Artificial Analysis ranks it fifth in its speech arena at 1278 Elo for $16.50 per million characters; Eleven v4 Turbo leads at 1335 for $40, with Eleven v4 second at $80. Flash-Lite TTS is eighth at 1251 for $11.
The price advantage is temporary. Google's pricing page holds current rates only to 31 December 2026, and every line doubles on 1 January, which puts Flash TTS near $33 per million characters on Artificial Analysis's basis and narrows the gap to Eleven v4 Turbo to a few dollars. Alibaba's Qwen-Audio-3.1-TTS-Plus already ranks above it at $19.30.
Google's launch post leans on Hume AI's benchmarks, where it says Flash TTS ranks first for voice design (71.4) and overall quality. Those results are cited by Google rather than reproduced independently, and Artificial Analysis's blind votes put four models ahead of it.
The engineering limits show up in dialogue. One request handles two speakers with prebuilt voices; designed or replicated voices must be generated turn by turn and stitched together as raw audio. Stored custom voices are capped at 200 per project and expire a year after last use, so a games studio with a large cast needs its own record of every voice prompt.
Where it earns its place
An audiobook producer narrating 200 titles a year of about 10 hours each
Narrate backlist titles in one consistent designed voice
Design the narrator once, store it, and generate chapters through the Batch API: 2,000 hours is about $810 at batch rates now and about $1,620 after January.
A voice-agent engineer building on LiveKit or Pipecat
Cheap speech output for a phone agent
Use Flash-Lite TTS with streaming and request audio/mulaw at 8 kHz for telephony. Output costs about $0.54 per hour of speech now and $1.08 from January.
A localisation agency dubbing training videos into 30 languages
Dub at volume without a voice actor per language
Flash-Lite covers 100+ languages; at $0.54 an hour, 1,000 hours of dubbed audio is about $540 before January. Voice replication is not offered for UK or EEA users in AI Studio, so plan on library or designed voices there.
Before you commit
- Setup
- An AI Studio API key and google-genai 2.25.0+ (Python) or @google/genai 2.24.0+ (JavaScript); non-streaming calls return 24 kHz mono WAV.
- Integrations
- Gemini API with a Voices endpoint (/v1beta/voices) for design and replication; Google lists Agora, LiveKit, Pipecat and Vercel as platforms, and Google Vids and Gemini Notebook for end users.
- Data and privacy
- Free-tier text and audio may be used to improve Google's products; paid-tier data is not. Apps serving the EEA, UK or Switzerland must use paid services. All audio is SynthID-watermarked.
- Cost at scale
- Audio is billed at 25 tokens a second: Flash TTS costs about $0.81 per hour now and $1.62 from January, Flash-Lite $0.54 and $1.08. Batch and Flex halve both.
- Limits
- Two speakers per request with prebuilt voices; 200 stored voices per project, kept for a year after last use; stateless replicated-voice keys expire after 7 days. Rate limits are shown only in AI Studio.
What's good
- Fifth in Artificial Analysis's speech arena at $16.50 per million characters
- Voice design from a text description
- Voice replication with a spoken consent check
- Flash-Lite at $11 per million characters for volume
- Batch halves the price
What's not
- Every price doubles on 1 January 2027
- Replication unavailable in the EEA, UK, Switzerland, India, Illinois and Texas
- Only two prebuilt-voice speakers per request
- Behind Eleven v4, Qwen-Audio 3.1 and Sonic 3.6 in blind votes
Best for
Developers producing long-form narration or multilingual dubbing who want near-top quality and can absorb January's price rise.
Alternatives
Gemini 3.8 Flash TTS: common questions
- Is Gemini 3.8 Flash TTS worth it?
- Fifth in blind votes at under half Eleven v4 Turbo's price, Gemini 3.8 Flash TTS is a bargain with an expiry date of 1 January. We rate it 4.5 out of 5.
- How much does Gemini 3.8 Flash TTS cost?
- Flash TTS: $0.50 per 1M text tokens in, $9 per 1M audio tokens out (about $0.81 per hour of audio) until 31 Dec 2026, doubling on 1 Jan 2027. Flash-Lite: $6 per 1M audio tokens out. Batch half price; free tier in AI Studio.
- What are the downsides of Gemini 3.8 Flash TTS?
- Every price doubles on 1 January 2027; replication unavailable in the EEA, UK, Switzerland, India, Illinois and Texas; only two prebuilt-voice speakers per request; behind Eleven v4, Qwen-Audio 3.1 and Sonic 3.6 in blind votes.
- Who is Gemini 3.8 Flash TTS best for?
- Developers producing long-form narration or multilingual dubbing who want near-top quality and can absorb January's price rise.
- How much does Gemini 3.8 Flash TTS cost for a team or in production?
- Audio is billed at 25 tokens a second: Flash TTS costs about $0.81 per hour now and $1.62 from January, Flash-Lite $0.54 and $1.08. Batch and Flex halve both.
- Is Gemini 3.8 Flash TTS safe for company data?
- Free-tier text and audio may be used to improve Google's products; paid-tier data is not. Apps serving the EEA, UK or Switzerland must use paid services. All audio is SynthID-watermarked.
- What are the best Gemini 3.8 Flash TTS alternatives?
- The closest alternatives we have reviewed are Eleven v4 and v4 Turbo and Cartesia Sonic 3.6.
Answers checked Oct 22, 2026.
Make Gemini 3.8 Flash TTS? Show this rating on your site
