Gemini 3.8 Flash TTS and Flash‑Lite TTS are the newest voice‑generation models in the Gemini family. They support more than 100 languages and dialects and ship with a library of over 2,000 production‑ready voices. Flash TTS is built for deep creative direction, allowing developers to craft brand‑new characters from natural‑language prompts and control every line of speech. Flash‑Lite TTS is optimized for high‑volume, cost‑effective scaling, ideal for dubbing, podcasts, and expressive voice agents.
Key capabilities
- Custom voice creation: Describe role, accent, emotion, etc. in plain text; the model synthesizes a voice from scratch or from a 30‑second reference clip. The library can grow from 30 original voices to an unlimited collection.
- Voice replication: With a short sample and explicit consent verification, the system produces a copy that carries a SynthID watermark and C2PA credentials, protecting both developers and talent.
- Line‑by‑line script control: In Google AI Studio, Gemini API, Gemini Notebook, or Google Vids, you can embed directives such as
<laughs>,|mhm|, or pacing cues to shape pacing, emotion, and back‑channeling. - Long‑form generation: Maintains consistent timbre and natural pacing over hours of audio, suitable for audiobooks and podcasts without speaker drift.
- Two‑speaker scene staging: A single script can drive a dialogue between two distinct voices with natural turn‑taking.
- Safety features: Every output is watermarked with an imperceptible SynthID tag; voice cloning requires a verbal consent recording; a detailed model card outlines responsible use.
Performance: On Hume AI’s Voice Design Benchmark, Flash TTS scores 71.4 overall (rank 1) and 60.8 on accent modeling (rank 1). Flash‑Lite TTS follows closely. In blind human preference tests on Voice Arena, both models rank top across major languages such as Japanese, Brazilian Portuguese, Vietnamese, Modern Standard Arabic, Mexican Spanish, and Hindi.
How to use
- Developers: Call
gemini.tts.flashorgemini.tts.flash_litevia the Gemini API or Google AI Studio. - Enterprises: Availability coming soon through Gemini Enterprise API.
- General users: Accessible today in Gemini Notebook and Google Vids.
Partners like Figma, HeyGen, Linguana, Wondercraft, 99.co, and Ollang are already integrating the models for global dubbing, nuanced regional localization, and scalable conversational agents.
Review