What is Qwen3-TTS?
Alibaba Qwen's open TTS models for voice cloning, voice design and preset voices in 10 languages.
Who is Qwen3-TTS best for?
Qwen3-TTS is best for developers who want open speech models they can self-host to clone a voice or design one from text.
Voice cloning
The Base models copy a voice from about 3 seconds of audio and can serve as a starting point for fine-tuning.
Voices from a description
VoiceDesign 1.7B builds a new voice from a plain-language description of how it should sound.
Real-time speech apps
Streaming output starts after a single character of input, with end-to-end latency as low as 97 ms.
Products in several languages
It speaks Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish and Italian, plus dialect voice profiles.
How much does Qwen3-TTS cost?
Qwen3-TTS is free and open source under Apache 2.0, and Alibaba Cloud also serves it as a real-time API through Model Studio (DashScope).
- Apache 2.0 code and weights
- 0.6B and 1.7B models on Hugging Face and ModelScope
- Fine-tuning and vLLM support
Alibaba Cloud also offers real-time Qwen3-TTS APIs for custom voices, voice cloning and voice design through Model Studio (DashScope), documented and billed on Alibaba Cloud's own pages. Full pricing on github.com. We checked these prices on .
How do readers rate Qwen3-TTS?
No one has rated Qwen3-TTS yet. Be the first.
Sign in to rate. One rating per account.
Is Qwen3-TTS open source?
Yes. Qwen3-TTS's code is public under the Apache 2.0 licence, so you can run it on your own server.
On GitHub
14,565stars
- Last commit
- 17 Mar
- Repository age
- 10 months
- Licence
- Apache 2.0
- Self-hosted
- Yes
- Repository
- QwenLM/Qwen3-TTS
What are the best Qwen3-TTS alternatives?
Qwen3-TTS is an open source alternative to MAI-Voice-2, MiMo-V2.5-TTS, Fish Audio S1 and 3 more.
Microsoft AI's expressive text-to-speech model for 10+ languages, with voice prompting from a short clip.
Xiaomi's text-to-speech models with stock voices, voice design from a prompt and voice cloning, via API.
Expressive text-to-speech and voice cloning in 13 languages.
Fish Audio's open-weights text-to-speech model with inline emotion tags, multi-speaker dialogue and 80+ languages.
ElevenLabs' expressive speech model for voiceovers and dialogue in 90+ languages.
ElevenLabs' Voice Design V3 creates custom voices from a text prompt.
Questions about Qwen3-TTS
What is Qwen3-TTS?
Alibaba Qwen's open TTS models for voice cloning, voice design and preset voices in 10 languages.
Who is Qwen3-TTS for?
Qwen3-TTS is best for developers who want open speech models they can self-host to clone a voice or design one from text.
Is Qwen3-TTS free?
Yes. Qwen3-TTS has a free plan: apache 2.0 code and weights.
Is Qwen3-TTS open source?
Yes. Qwen3-TTS's code is public under the Apache 2.0 licence, so you can run it on your own server.
What are the best Qwen3-TTS alternatives?
The closest alternatives on Recaply are MAI-Voice-2, MiMo-V2.5-TTS, Fish Audio S1, Fish Audio S2 Pro, Eleven v4 and ElevenLabs Voice Design V3.
Where can I try Qwen3-TTS?
On its official website, github.com.
How much audio does Qwen3-TTS need to clone a voice?
About 3 seconds. The Qwen3-TTS Base models, in 0.6B and 1.7B sizes, clone a voice from a short clip of user audio, and Qwen says they can also be fine-tuned into other models.







