Google Launches Gemini 3.8 Flash TTS Models With Custom Voice Creation and Control

Google has rolled out two new text-to-speech models, Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, with tools for creating and controlling custom voices. The models can generate new voices from text prompts, replicate voices from 30-second samples and produce dialogue with different tones, pacing and performance cues. Google is also adding support for multi-speaker conversations, long-form audio and more than 100 languages and dialects. The two models are now available across several Google products and developer tools.

Gemini 3.8 Flash TTS, Flash-Lite TTS Availability

According to a Google blog post, Gemini 3.8 Flash TTS is now available through the Gemini API and Google AI Studio for developers. It is also rolling out to Gemini Notebook users. API access for enterprises through Gemini Enterprise is planned for a later date.

Gemini 3.8 Flash-Lite TTS is also available through the Gemini API and Google AI Studio. Google is bringing the model to Google Vids, while its Gemini Enterprise API rollout will follow later.

The two models have different use cases. Flash TTS is aimed at applications that need detailed control over voice creation and delivery, while Flash-Lite TTS is designed for higher-volume tasks such as dubbing, audio generation and voice agents.

What Gemini 3.8 TTS Can Do

Gemini 3.8 Flash TTS can create voices from natural language descriptions. Users can define characteristics such as accent, role and vocal style, with support for more than 100 languages and dialects. Google also provides access to more than 2,000 ready-made voices.

The model can recreate a voice from a 30-second audio sample when the user has permission to use it. Google requires a consent recording from the voice owner before allowing replication. Generated audio also carries a SynthID watermark.

Both models let users control individual lines of dialogue with instructions for delivery, tone and pacing. They can generate conversations between two speakers from a single script and maintain separate voices during the exchange. Scripts can also include non-verbal sounds such as laughs, sighs and gasps, along with conversational responses.

Gemini 3.8 Flash TTS can handle longer audio projects while maintaining the selected voice across the recording. This could be useful for formats such as podcasts and audiobooks.

Custom Voice Features Coming Later

Google is also working on a voice remixing feature for Gemini 3.8 Flash TTS. It will allow users to start with an existing voice and modify characteristics such as pitch, timbre, pace and accent through text prompts.

Google AI Studio has also received an audio playground where developers can create custom voices or replicate an authorised voice before using them in a dual-speaker screenplay editor.

Google says the new models improve long-form generation and two-speaker script control compared with Gemini 3.1 Flash TTS. It also reported strong results for the models in its own evaluations, including Hume AI's Voice Design Benchmark and Overall Quality Index, as well as blind human preference tests covering languages such as Japanese, Hindi, Brazilian Portuguese and Mexican Spanish.

The Gemini 3.8 TTS models join Google's existing Gemini Audio offerings, which include Gemini 3.5 Live Translate, Gemini 3.5 Transcribe, Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking.