Qwen3-TTS robustness improvement: more stable under noisy text and complex inputs

Qwen3-TTS open source release: 12Hz high-compression tokenizer + 3 seconds of tone clone How to play

1. Abstract Qwen3-TTS is a family of open-source text-to-speech (TTS) models from the Qwen team, including VoiceDesign (to generate new voices from te...