Qwen3-TTS is a family of editions that each do one job well. Avocado recognises which one you have and shows only what it can do.
Presets
The Presets edition has nine ready voices, including two that speak Beijing and Sichuan dialects.
Clone
The Base editions speak with a voice you clone. Give the clip’s words too and you get a closer clone; Avocado saves it as a ready voice almost instantly. Only clone voices you have permission to use.
Design
With the Voice Design edition, a description becomes the voice. Save it and Avocado renders it once, so the same voice also works on the faster Base editions.
Train
The 0.6B Base edition can be taught a voice from your own recordings. Avocado keeps the best result it finds and stops early when more steps won’t help.
Licence
Qwen3-TTS’s licence is set by Alibaba Qwen. Read it on the Qwen3-TTS model page before you publish anything you make.