TOOLORA / Local reading edition
Qwen3-TTS
Choose VoiceDesign, CustomVoice or Base checkpoints for described voices, preset-voice control or reference cloning across ten languages.
Overview
Qwen3-TTS separates tasks by checkpoint: VoiceDesign creates voices from descriptions, CustomVoice uses preset speakers, and Base reproduces reference voices and supports fine-tuning. Released models cover ten languages, but instruction control differs between sizes and variants. Python, local WebUI, streaming and non-streaming workflows are documented. Hosted APIs require separate account and service configuration. Main code and the reviewed official model card declare Apache 2.0.
- Company country / region
- China · Company headquarters
Official sources and references
- Official websitehttps://github.com/QwenLM/Qwen3-TTS
- Documentationhttps://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign
Descriptions summarize public information; consult official websites for features, pricing and regional availability.