What Gemini 3.1 Flash TTS is good at — and what it's not

Gemini 3.1 Flash TTS is Google's expressive text-to-speech model, built on the Gemini speech lineage. Its two defining strengths are a library of 30 named voices — Kore, Puck, Zephyr, Achernar, and more, spanning a wide range of tones and characters — and a separate style instruction field that lets you direct the delivery. The same script can be read upbeat, calm, or dramatic depending on the instruction, which gives finer control than picking a voice alone.

Pricing is per 1,000 characters, so cost scales cleanly with script length — predictable for everything from a short voiceover to a full narration. It's a natural pairing for talking-avatar work: generate the voice here, then drive a model like Kling Avatar V2 with it.

Where it sits among ZOOOP's voice models: Multilingual V3 is ElevenLabs' flagship with deep voice control; Qwen3-TTS and Inworld TTS lead on multilingual coverage and value. Gemini 3.1 Flash TTS's sweet spot is expressive, style-directed narration with Google's voices.

A reasonable mental model: default to Gemini 3.1 Flash TTS when you want expressive narration with explicit style control, and switch to Multilingual V3 for ElevenLabs' voice library or Inworld/Qwen for broad multilingual coverage.