Google Launches Gemini 3.8 Flash TTS and Flash-Lite TTS

Google this week released two text-to-speech models, Gemini 3.8 Flash TTS and Gemini 3.8 Flash-Lite TTS, calling them its most expressive audio generation models to date. They rank first and second on Hume's voice quality benchmark and are available now in Google AI Studio, with voice cloning requiring only a 30-second sample.

Confirmed

• Positioning: Flash TTS targets creative production, while Flash-Lite TTS targets efficiency, scale, and low-cost mass speech synthesis.
• Voice design: custom voices with distinct accents and personalities can be designed via text prompts.
• Fine-grained control: tone and pacing, including whispering, speech rate, and emotion.
• Voice cloning from just 30 seconds of audio.
• Native multi-speaker conversation with interruptions and laughter.
• Use cases include voice design, character dubbing, and dubbing translation; developers can integrate via official SDKs.

Why it matters

• Cloning门槛 at 30 seconds plus prompt-based voice design sharply lowers production costs for personalized voice content, and may intensify safety and compliance debates around voice impersonation.
• Sweeping the top two spots on the third-party Hume benchmark puts direct competitive pressure on rivals in speech generation quality.

2026-09-24 ~ 2026-09-26 · 5 related posts

• DeepMind ships Gemini 3.8 Flash TTS with native multi-speaker overlap and laughter — kastnerkyle · 2026-09-24
• Google's Gemini 3.8 Flash TTS: prompt-designed voices, cloning, and Java SDK — rseroter · 2026-09-25
• [source] Google launches Gemini 3.8 Flash TTS and Flash-Lite TTS, its most expressive audio models yet — CurieuxExplorer · 2026-09-25
• [source] Google ships Gemini 3.8 Flash TTS: voice cloning from 30 seconds, tops Hume benchmark — thursdai_pod · 2026-09-26
• [source] Google releases Gemini 3.8 Flash TTS and Flash-Lite TTS speech models — CurieuxExplorer · 2026-09-26