Gemini 3.8 Flash TTS: Google's Text-to-Speech Model, Voices, Languages, Limits and Pricing

What Gemini 3.8 Flash TTS is

Gemini 3.8 Flash TTS is a text-to-speech model from Google, generally available on the Gemini API since 22 September 2026 under the model code gemini-3.8-flash-tts. Google describes it as its flagship creative text-to-speech model, built for studio-grade voice fidelity, expressive acting, authentic regional accents and stability over long, multi-turn audio. It takes text only as input and returns audio only. Google positions it for audiobooks, studio narration, complex multi-speaker dialogue, heavy vocal-burst acting, difficult pronunciation and regional dialects.

Flash TTS and Flash-Lite TTS

Google released two Gemini 3.8 speech models that share the same API schema and prompting structure. Gemini 3.8 Flash TTS aims at maximum voice fidelity, acting nuance and dialect coverage and supports 130 languages. Gemini 3.8 Flash-Lite TTS aims at high throughput, low latency and lower cost, supports 101 languages, and is the recommended replacement for gemini-3.1-flash-tts-preview; Google suggests it for high-volume production, real-time voice agents, read-aloud features and everyday single-speaker generation. Text-to-speech through the Gemini API is separate from the Live API, which handles interactive, unstructured conversations.

Controlling delivery

The 3.8 models treat input text strictly as a verbatim transcript, so instructions written into the text, such as a line beginning with a speaking style or a speaker label, may be read aloud. Sustained delivery instructions and speaker labels go instead into structured speech_metadata attached to each turn, and short vocal events use angle-bracket inline tags such as laugh, sigh or short pause. Google states that the model keeps a consistent voice identity, timbre, volume and room tone across long dialogues and multi-minute narration. Teams migrating from gemini-3.1-flash-tts-preview or earlier Gemini speech models need to move to this schema.

Voices and limits

Voices can come from the prebuilt set, the Extended Voice Library, custom personas made with voice design, voice replication, or voice remixing, and can also be designed in Google AI Studio. A single multi-speaker request supports up to two speakers using prebuilt voices; dialogue between custom or replicated voices is made by generating each turn separately and joining the 24kHz PCM audio. Stored custom voices are limited to 200 per project with a one-year lifetime, and stateless voice keys last seven days. Requests return WAV by default, and raw PCM can be requested instead.

Pricing

Gemini 3.8 Flash TTS is free of charge on the Gemini API free tier, where Google may use the data to improve its products. On the paid tier, through 31 December 2026, text input costs $0.50 per million tokens and audio output $9.00 per million tokens, which Google gives as about $0.00225 per 10 seconds of audio. From 1 January 2027 the prices double to $1.00 for text input and $18.00 for audio output, about $0.0045 per 10 seconds. Batch and Flex processing cost half the Standard price: $0.25 input and $4.50 output through the end of 2026.

Official Resources