Alibaba unveils beta of AI music model that generates full songs from text

What was released

Alibaba Group has unveiled a beta version of "HappyShrimp," an AI music generation model that produces a complete song from a single line of text. Reportedly, when a user types a sentence describing a desired mood or theme, the model generates a finished track complete with lyrics, composition, and arrangement. The announcement comes as part of Alibaba's broader push to expand its generative AI business, according to Bloomberg Technology.

Where to try it

HappyShrimp is usable directly on the web as soon as it launches. After logging in at the official HappyShrimp site, users can describe the song they want in an input box (up to 5,000 characters), and two tracks are generated at once. A toggle lets users choose between "Smart Lyrics," which automatically adds lyrics, and "Instrumental," which produces music without vocals.

The site states that logging in daily during the launch period earns free credits. No separate pricing has been posted yet (as of August 19, 2026). The homepage showcases user-created tracks in "Trending Now," "Global Top," and "Fresh Drops" playlists, and Asia-focused genre presets — such as K-pop dance tracks or Asian pop-rock ballads — stand out. The site's product lineup also lists "Happy Horse" and "Happy Oyster" alongside HappyShrimp.

What this means

Services that produce a finished song from text alone are already an area opened up by Suno and Udio. Suno has been credited with dramatically lowering the barrier to entry for music creation by generating tracks combining vocals and instrumentation from just a short description. How deeply this technology has already permeated real-world use was underscored by a recent case: string instrument parts maker D'Addario denied for about two weeks that AI music tools were used in a promotional video — blaming low-quality exports or mastering plugins — before eventually admitting to using Suno.

Over recent months, Alibaba has expanded its generative AI portfolio in multiple directions — including coding and image generation — through its Qwen line of models. When a major tech company that already has text-to-image and text-to-video capabilities also secures text-to-music, it can bundle content-creation tools across an ecosystem spanning search, shopping, and cloud services. The pattern of music generation being absorbed into platform features rather than existing as a standalone app echoes similar moves already seen with other tech giants' image- and video-generation tools.

What changes now

The mere fact that a major Chinese tech company has entered the text-to-music generation space directly broadens the competitive landscape of this market. As large platform operators like Alibaba join a market previously centered on Suno and Udio, the practice of directly generating music for use in advertising and marketing is likely to become even more common. The launch period, with its free credits, is also a window for domestic creators to directly compare output quality. To try it yourself, go to HappyShrimp, log in, type a single line describing the song you want, and give it a try.