- Published on
Gemini 3.8 TTS Playground
- Authors
- Name
- DeepHorus
On 23 September 2026, Google released two Gemini text-to-speech models: gemini-3.8-flash-tts and gemini-3.8-flash-lite-tts. The launch broadens Google’s speech-generation lineup with two options built for turning text into audio. Their importance lies in combining voice selection, custom voices and multi-character dialogue controls.
Both models offer a library of more than 2,000 voices. Developers can assign different voice-style instructions and define a complete conversation among multiple characters. These features could simplify projects that previously required separately produced and synchronized speech clips.
Users can also create a custom voice from a 30-second audio sample. The sample must be of the user’s own voice or one the user has rights to use. This enables tailored narration while making consent and licensing central requirements.
A developer built a bring-your-own-key playground for the release with GPT-6 Astra. Its interface uses the Gemini API’s open CORS policy to support browser interaction. The project shows how third parties can build direct web tools around Google’s speech models.
A demo featured two pelicans debating whether they should move to Pacifica Pier. Claude 4.5 Opus wrote the script and generated a URL that rendered the exchange through the tool. The pelicans used separate voices and voice-style instructions, illustrating the API’s multi-speaker workflow.
The reported test produced 1 minute and 18 seconds of audio in about 20 seconds using Gemini 3.8 Flash TTS. The render cost 2.74 cents and did not use the cheaper Flash-Lite model. The figures come from one demonstration rather than a broad benchmark of speed, cost or quality.
The release targets narration, character dialogue and authorized voice customization. It pairs a large voice catalog with API-level control over speaker identity, style and conversational structure. Wider adoption may depend on how consistently the tools balance audio quality, affordability, rights compliance and ease of use.
