Gemini 3.1 Flash TTS 发布：自然语言 audio tags 可精细控制语速、音色与场景，Artificial Analysis TTS 榜 Elo 1211，支持 70+ 语言
Gemini 3.1 Flash TTS: the next generation of expressive AI speech
精选评分 61
导读
Google DeepMind 发布 Gemini 3.1 Flash TTS，新一代文本转语音模型，在可控性、表现力与音质上均有提升。今日起铺开：开发者可在 Gemini API 与 Google AI Studio 预览，企业在 Vertex AI 预览，Workspace 用户经 Google Vids 使用。官方称其音质更自然、更具表现力——在 Artificial Analysis TTS 排行榜（汇总数千次人类盲评）拿到 1211 的 Elo 分，并被 Artificial Analysis 归入高音质与低成本的具吸引力象限。模型原生支持多说话人对话、70+ 语言，并引入 audio tags：在文本输入里嵌入自然语言指令来精确操纵语音风格、节奏与演绎，配场景调度、说话人级 Audio Profiles 与 Director notes 等控制。
推荐理由
把语音合成从念对字推进到可导演：自然语言标签控制语速音色与场景，且给出第三方榜单的量化位次。
主体与标签
Google DeepMind#产品更新#模型发布#语音#多模态
正文
Our newest audio model introduces granular audio tags that give you precise control to direct AI speech for expressive audio generation.
Member of the Technical Staff
Principal Research Engineer on behalf of the Gemini team
Your browser does not support the audio element.
This content is generated by Google AI. Generative AI is experimental
Today, we’re introducing Gemini 3.1 Flash TTS, the latest text-to-speech model that delivers improved controllability, expressivity and quality — empowering developers, enterprises and everyday users to build the next generation of AI-speech applications.
Starting today, 3.1 Flash TTS is rolling out:
- For developers in preview via the Gemini API and Google AI Studio
- For enterprises in preview on Vertex AI
- For Workspace users via Google Vids
Improved speech quality and controllability
We’ve improved the overall speech quality of Gemini 3.1 Flash TTS, making it our most natural and expressive model to date. On the Artificial Analysis TTS leaderboard , a benchmark that captures thousands of blind human preferences, 3.1 Flash TTS achieved an impressive Elo score of 1,211.
Artificial Analysis has also positioned Gemini 3.1 Flash TTS within its “ most attractive quadrant ” for its ideal blend of high-quality speech generation and low cost. The model stands out further with native multi-speaker dialogue, support for 70+ languages, and granular creative control via natural language.
New audio tags for more expressive speech generation
3.1 Flash TTS also introduces audio tags — an intuitive way to control vocal style, pace and delivery. By embedding natural language commands directly into the text input, you can steer AI-speech output with improved levels of granularity.
3.1 Flash TTS enables enterprises to utilize audio tags within Vertex AI, empowering the next generation of enterprise applications.
3.1 Flash TTS lets you use audio tags to experiment with speech expressivity, pacing and delivery.
3.1 Flash TTS lets you integrate expressive speech capabilities, turning a standard weather app into an engaging experience.
3.1 Flash TTS supports a range of audio tags to add nuanced, engaging delivery to a synonym word hunt app.
You can start experimenting with these audio tags along with other updates to the developer experience in Google AI Studio with configurable controls that place the developer in the “director’s chair”:
- Scene direction: Set the stage by defining the environment and providing specific dialogue instructions. This world-building context helps characters remain “in-character” and react to one another naturally across multiple turns.
- Speaker-level specificity: Cast characters using unique Audio Profiles, then specify Director’s Notes to toggle pace, tone and accent. Using inline tags , speakers can pivot from these high-level settings to change expression mid-sentence.
- Seamless export: Once the performance is perfected, these exact parameters can be exported as Gemini API code to ensure consistent, recognizable voices across various projects and platforms.
With these new configurations, developers can enhance precision for specific scenarios, creating memorable characters and immersive audio experiences.
Built for global scale
Gemini 3.1 Flash TTS delivers high-fidelity speech and more precise control across more than 70 languages. These core optimizations bring advanced style, pacing and accent control to major markets — helping developers create localized, expressive speech experiences for users at global scale.
Early developer and enterprise testers are already seeing the impact of 3.1 Flash TTS, highlighting its impressive controllability and expressivity. They’ve told us how audio tags provide a new level of creative precision, transforming simple text into a high-fidelity vocal performance.
Watermarked with SynthID
All audio generated by Gemini 3.1 Flash TTS is watermarked with SynthID. This imperceptible watermark is interwoven directly into the audio output, allowing the reliable detection of AI-generated content to help prevent misinformation. For more information on our approach to safety and responsibility, you can review the model card .