Alibaba Launches HappyShrimp 1.0: A Song That Can Be Sung in One Breath, Plus an AI-powered "Xiami Music"

According to the AI ​​Watch monitoring, Alibaba has officially released the AI music model HappyShrimp 1.0, Chinese name "快乐虾米" (Happy Shrimp). Users only need to describe an emotion, story, or music idea, and the model can directly complete the lyrics, composition, arrangement, and singing to generate a complete song.

HappyShrimp focuses on natural language understanding, without requiring users to first learn a bunch of music tags. For example, by entering "a letter to my recent graduate self" or "a song suitable for café background music," the model will decide on the style, instruments, vocals, and song structure. Complex prompts can also control BPM, key, instruments, vocal gender, singing style, and emotional progression.

The model adopts end-to-end full-song generation. The lyrics, melody, arrangement, and vocals are not generated separately and then pieced together but are uniformly planned for the entire song, with a focus on addressing common AI music issues such as robotic vocals, mismatched lyrics and melody, and loose song structures.

Alibaba has also simultaneously launched the eponymous AI Music Platform HappyShrimp. In addition to directly generating songs, users can also publish, listen to, and share other people's AI music works. Currently, the domestic and overseas web versions have been opened, while an independent app is still in preparation for launch.

The launch of HappyShrimp was accompanied by an announcement of a collaboration with the Taihe Music Group. New users can currently receive free credit points to experience the platform.