Qwen3 TTS TTS
Alibaba's 1.7B multilingual model offering both 9 preset voices and voice design from a plain-text description.
Wrap wanu malemba mu SSML tags kwa kuwongolera moyenera:
<speak><prosody rate="slow">Slow speech</prosody></speak>
Tags chosankhidwa chitsanzo amamvetsa - dinani kuti aphe mmodzi m'mawu anu pamene chimachitika:
Izi ndi njira yolemba malemba oyera, kotero ma tag ophatikizidwa amasiya kuganiziridwa. Kuti mupange ma tag ogwirizana ndi maganizo, gwiritsani ntchito njira yolemba malemba monga Orpheus kapena Bark.
Define custom pronunciations (word = pronunciation):
Za Qwen3 TTS
Qwen3-TTS is a 1.7-billion-parameter model from Alibaba's Qwen team, built on the Qwen3 backbone with a multi-codebook decoder. What sets it apart is its dual operating mode: you can pick from 9 preset speakers with emotion control, or switch to a voice-design mode where you describe the voice you want in natural language and the model synthesizes a matching one. It spans 10 languages — English, Chinese, Japanese, Korean, German, French, Russian, Portuguese, Spanish, and Italian — with high expressiveness and natural prosody. It handles up to 2,000 characters per request, making it a flexible choice for multilingual content that needs either consistent preset voices or bespoke custom ones.
Best kwa: Multilingual content with preset voices or custom voice design
Pezani zonse Qwen3 TTS maganizoPa mphindi
- Wopanga
- Alibaba (Qwen)
- License
- Apache 2.0
- Mtundu
- standard
- Kuyenda
- medium
- Kusintha kwa mawu
- Si
- Zilankhulo
- English, Chinese, Japanese, Korean, German, French, Russian, Portuguese, Spanish, Italian
- Max characters
- 2000