업데이트
What is Gemini 3.8 Flash Lite TTS?
Gemini 3.8 Flash Lite TTS is Google's high-volume text-to-speech model from the Gemini 3.8 family, released on September 23, 2026. It is built for high-volume speech: narration, dubbing, read-aloud features, and voice agents. On UnificAlly it runs as google/gemini-3.8-flash-lite-tts and is billed per second of audio delivered.
The quality holds up. On the Artificial Analysis Text to Speech leaderboard it ranks 6th with an Elo of 1241, 38 points above Gemini 3.1 Flash TTS, the model Google names it as a replacement for.
What's new in Gemini 3.8 Flash Lite TTS
Top 6 on the Artificial Analysis TTS leaderboard
Gemini 3.8 Flash Lite TTS scores an Elo of 1241, 6th overall, between Inworld Realtime TTS-2 (1246) and Simba 3.2 (1240). It beats Gemini 3.1 Flash TTS (1203) and trails only its sibling, Gemini 3.8 Flash TTS (1267), among Gemini voices.
Built for volume, same request as Flash
Google tunes Flash Lite for throughput and low latency on bulk narration, dubbing, and voice agents. It takes the same request shape as Flash, so you can run volume jobs here and keep Flash for the lines that need acting.
Vocal bursts. The same angle-bracket sounds as Flash, like <laugh>, <sigh>, and <short pause>, placed exactly where they should play.
Two-speaker dialogue. One request renders a conversation between two prebuilt voices.
100+ languages. Language is detected from the text automatically.
The 30 Gemini voices. Kore, Puck, Charon, and the rest of the curated set, plus Google's extended voice library.
What Gemini 3.8 Flash Lite TTS is best for
Dubbing at scale
Large batches of translated lines rendered with one consistent voice per speaker.
Voice agents
Spoken replies for support and booking agents that run all day.
Read-aloud features
Article, email, and document playback inside an app.
E-learning narration
Course modules and product tours voiced in 100+ languages.
Notifications and prompts
Short system lines for IVR menus, alerts, and in-app guidance.
Gemini 3.8 Flash Lite TTS use cases
Add a "listen to this article" button that voices every post the moment it is published. Dub a video library into a dozen languages from translated scripts, with one voice per speaker kept across every episode. Give a support agent a spoken voice for phone and in-app calls, where fast replies matter more than acting. Voice short product walkthroughs and onboarding steps for each market you ship to.
Gemini 3.8 Flash Lite TTS limitations
Google lists Gemini 3.8 Flash Lite TTS as Preview on its enterprise API, so voices and limits can still change. It covers 100+ languages against 130+ for Flash, and it ranks below Flash on blind listening tests (1241 against 1267 Elo). Dialogue tops out at two speakers per request, and each request takes up to 8,192 tokens of text.
When to use Gemini 3.8 Flash Lite TTS
Use it as the default for anything that runs at volume, and move only the lines that need real acting up to Gemini 3.8 Flash TTS. Since both share one request shape, switching is a change of model ID.
Gemini 3.8 Flash Lite TTS API examples
curl -X POST https://api.unifically.com/v1/tasks -H "Content-Type: application/json" -H "Authorization: Bearer YOUR_API_KEY" -d '{
"model": "google/gemini-3.8-flash-lite-tts",
"input": {
"text": "Your order has shipped and arrives tomorrow.",
"voice": "Puck",
"response_format": "mp3"
}
}'
Poll GET /v1/tasks/{task_id} until status is completed, then read output.audio_url. Full parameter reference, voices, and dialogue examples live at docs.unifically.com.
FAQ
Gemini 3.8 Flash Lite TTS에 대해 자주 묻는 질문
Send a POST /v1/tasks request with model set to google/gemini-3.8-flash-lite-tts and your text and voice in input. The request body matches Gemini 3.8 Flash TTS, so switching between them is a change of model ID.
Google released Gemini 3.8 Flash-Lite TTS on September 23, 2026, together with Gemini 3.8 Flash TTS. Both launched on the Gemini API and in Google AI Studio.
It sits 6th on the Artificial Analysis Text to Speech leaderboard with an Elo of 1241, just behind Inworld Realtime TTS-2 at 1246 and ahead of Simba 3.2 at 1240. That is 38 points above Gemini 3.1 Flash TTS at 1203, the model Google positions it to replace.
Each request takes up to 8,192 input tokens of text, and Google serves up to 16,384 audio output tokens. Audio runs at about 25 tokens per second, so one request covers roughly ten minutes of speech.
Yes. It takes the same angle-bracket vocal bursts as Flash, like <laugh> or <short pause>, and renders two-speaker dialogue with prebuilt voices in one request.
More than 100 languages with automatic detection, against 130+ for Gemini 3.8 Flash TTS. The 30 curated Gemini voices, such as Kore and Puck, are available on both.
Pick Flash Lite for high-volume and latency-sensitive work, like dubbing, read-aloud features, and voice agents. Pick Flash when acting quality or the widest language coverage matters more.
Gemini 3.8 Flash Lite TTS 가이드 및 비교
이 모델을 실행하기 전에 가격, 제한 사항, 테스트된 결과물을 확인하세요.
