HappyShrimp

happyShrimp is an AI music generation model from Alibaba's ATH unit that turns a plAIn-language description of a mood, a story or a memory into a finished song. Instead of chaining separate systems for lyrics, melody, arrangement and singing, it plans the whole track in one pass, so words, instrumentation and vocals stay aligned. The current release runs in a desktop browser, with a mobile app described as still in preparation.

What it does

You describe what you want in everyday language rather than in production jargon. A Prompt can specify an EMOtional tone, a nARRative theme, a genre, a decade, a vocal type, a tempo or an instrumentation preference, and the model builds a complete track around that intent. The company positions the model as understanding cultural context and emotional intent, not just style labels, so requests like an eerie monologue from an abandoned ghost or a rainy-afternoon cafe atmosphere are treated as creative direction rather than keyword matching. Users can also supply their own lyrics or a reference track as a starting point.

Key capabilities

• text-to-song generation: one description produces lyrics, melody, arrangement and a sung vocal in a single ouTPUt.

• Lyric assistance: drafts, rewrites or restyles lyrics around a theme, story or set of keywords, generated in step with the music.

• Arrangement control: accepts direction on genre, era, vocal gender, singing style, key, tempo and instrumentation.

• voice replication: upload a reference audio Sample and APPly that timbre to a generated song, subject to authorization for the voice used.

• Reference-based creation: vibe referencing and remix-style reinterpretation of an uploaded track while keeping its core feel.

• Style conveRSIon: reshape an existing audio piece into another genre, such as a folk arrangement moving toward rock or electronic.

• Instrumental output: vocal-free tracks and separated vocal and backing stems for vIDEo, games or podcast use.

Who it is for

• Everyday users who want a Personal song built around a memory or feeling without any music training.

• Content creators needing original background music or theme tracks for short video, podcasts and vlogs.

• Video and game developers looking for fast, low-cost scoring and ambience.

• Songwriters using it for demos, style exploration and idea generation.

• marketing teams, educators and audio publishers who need custom music quickly.

How to get started

1. Open the web app and register an account; new users are typically granted a Free credit allowance.

2. Write a description covering mood, story, genre, vocals and tempo, or paste in lyrics.

3. Optionally attach a reference track or a voice sample.

4. Generate, listen, then iterate with Remix or reference-based variations.

5. Download the result; lossless export and commercial rights are tied to higher plan tiers.

Pricing and access caveat

Access follows a credit model combined with paid subscription tiers. Free accounts get a starter allowance and a liMITed feature set, while paid tiers add larger credit quotas, higher concurrency, priority queues and, at the top level, lossless downloads and commercial licensing. Generation cost per song is widely reported as very low, but credit consumption, plan names and pricing may change while the product is still new, so confirm current terms on the official site before committing. Free-tier output is Generally restricted for commercial use, Voice Cloning requires permission from the voice owner, and generated lyrics may fall short of Professional songwriting.