Qwen has open-sourced Qwen3-TTS in a public repository, shipping high-fidelity text-to-speech alongside voice cloning and “describe-a-voice” style voice design controls.