Gemini 3.1 Flash TTS

2026-04-15 · Publisher : Google · Domain : Artificial Intelligence

Google lance Gemini 3.1 Flash TTS en préversion, un modèle de synthèse vocale expressif et pilotable par balises audio, avec dialogue multi-locuteurs et plus de 70 langues.

What changes

  • Natural-language audio tags to control style, pace and delivery.
  • Native multi-speaker dialogue and per-character audio profiles in AI Studio.
  • More than 70 languages with control of style, pace and accent.
  • Elo of 1,211 on the Artificial Analysis TTS leaderboard (blind human preferences).
  • Audio output marked with SynthID.

Google launches Gemini 3.1 Flash TTS in preview, an expressive text-to-speech model steerable with audio tags, with multi-speaker dialogue and more than 70 languages. ### Specifications - Availability: Gemini API and AI Studio (preview), Vertex AI (preview), Google Vids. - Identifier: gemini-3.1-flash-tts-preview. ### Pricing | Item | Standard (USD / 1M tokens) | Batch | |---|---|---| | Input | 1.00 | 0.50 | | Output (audio) | 20.00 | 10.00 | Pricing taken from the current pricing page. ### Benchmarks | Benchmark (scores published by the vendor) | 3.1 Flash TTS | |---|---| | Artificial Analysis TTS (Elo) | 1,211 | ### Key takeaways A steerable synthetic voice model for voice applications and audio content; it was followed in September by the 3.8 TTS family.

Links

Tags : Audio, Google, Gemini, TTS, Voix