September 23, 2026
An app can speak 130 languages: Gemini TTS leaves preview
On September 22, Google moved Gemini 3.8 Flash TTS and Flash-Lite TTS out of preview. Flash TTS supports 130 languages, Flash-Lite TTS supports 101, and the models determine the text language themselves. Through December 31, Flash costs $0.50 per million input tokens and $9 per million output audio tokens, while Flash-Lite costs $0.50 and $6.
Previously, fast speech synthesis ran on `gemini-3.1-flash-tts-preview`. Gemini now offers Flash-Lite for high request volumes and retains the same API pattern with the same prompts. During migration, persistent delivery instructions move to `speech_metadata`, and the response now arrives as a WAV file with a RIFF header instead of raw PCM.
In Google AI Studio Audio Playground, you can create a voice or replicate your own, then assign lines to two speakers in the script editor. Replication requires only a 30-second sample, but Google requires a recording of the voice owner's consent. This feature is unavailable in Illinois, Texas, the EEA, the UK, Switzerland, and India.
For apps. Voices are created through the Voices endpoint `/v1beta/voices`, and audio through a POST request to `/v1beta/interactions`: the model, text, `response_format: audio`, and a voice in `generation_config.speech_config` immediately return `out.wav`. Flash TTS accepts text only and returns audio only, with a limit of 8 192 input and 16 384 output tokens.
On January 1, 2027, prices for both models will double: Flash to $1 and $18, Flash-Lite to $1 and $12 per million tokens.
Source
