Qwen3-TTS Releases Open-Source Voice-Cloning And Generation
Qwen releases the Qwen3-TTS family of open-source text-to-speech models, publishing tokenizers and models under the Apache 2.0 license. The models, trained on over 5 million hours across 10 languages, support three-second voice cloning, description-based control, streaming real-time synthesis via a dual-track LM, and state-of-the-art results on multilingual and long-speech benchmarks. Hugging Face hosts 0.6B (2.52GB) and 1.7B (4.54GB) variants with a browser demo that enables voice cloning.
Key Points
- 1Releases Qwen3-TTS family trained on over 5 million hours across 10 languages
- 2Provides state-of-the-art three-second voice cloning, description-based control, and streaming real-time synthesis
- 3Enables practitioners to run cloning locally or use Hugging Face browser demos with modest GPU
Scoring Rationale
Strong novelty and open-source release with large training data and demos; limited by incremental advances over prior voice-cloning systems.
Sources
Public references used for this report.
Practice interview problems based on real data
1,625 SQL & Python problems across 15 industry datasets — the exact type of data you work with.
Try 250 free problems