LectureRechercheRadarCadre d'investissement
Connexion / Inscription
← Retour à la lecture
ActualitésPrise rapide1 min de lectureRédaction DisconfirmAI

Google's new speech-synthesis models design voices from text descriptions

Voice generation adds description-based design and consent-gated cloning; promo pricing runs $0.54 per hour of audio until it doubles in 2027, with all capabilities vendor-claimed.

2026-09-24 01:39 GMT+8

Google released two speech generation models, Gemini 3.8 Flash TTS and Flash-Lite TTS, supporting more than 100 languages, rolling out through the Gemini API and Google AI Studio.

Flash TTS can design voices from scratch via text descriptions; cloning requires a 30-second sample plus a recorded consent statement, and generated audio carries a SynthID watermark. Both models support per-line stage directions, two-voice dialogue, and nonverbal sounds like laughter.

On pricing, an hour of audio output costs $0.81 with Flash TTS and $0.54 with Flash-Lite TTS through the end of 2026, doubling on January 1, 2027. The capabilities and benchmark lead are Google's own claims; the reporter's own tests found background whine and voice drift.

Sources

  • the-decoder.com↗
Connexion / Inscription
LectureRechercheRadarCadre d'investissement
siliconangle.com
↗
Continuer la recherche →