Alibaba's AI research team, Qwen (Tongyi Lab), released its speech synthesis AI, 'Qwen-Audio-3.0-TTS,' on July 20, 2026. By inputting reference audio and text, it can read any text aloud in any voice. The model supports 16 languages, including Japanese, English, and Chinese, and can handle 20 different Chinese dialects. It is available via API in two versions: 'Qwen-Audio-3.0-TTS-Flash,' which reduces latency to approximately 300 milliseconds, and 'Qwen-Audio-3.0-TTS-Plus,' which prioritizes quality. Qwen-Audio-3.0-TTS-Plus was rated as outperforming Gemini 3.1 Flash TTS in Artificial Analysis's blind tests and ranked first in the speech synthesis model performance rankings.