Via logodix.com
Fun-Music can produce complete tracks with lyrics, melody, arrangement, and vocals in both Chinese and English, though access is currently limited to China.
Alibaba Cloud just entered the AI music arena with Fun-Music, a model that takes a text prompt and spits out a fully formed song, complete with lyrics, melody, arrangement, and vocals.
The model is currently available in limited preview through Alibaba Cloud’s Model Studio, restricted to users in the China (Beijing) region who successfully apply for access. It supports both Chinese and English vocals, instrumental-only tracks, and outputs in MP3 or WAV formats.
What Fun-Music actually does #
The model works by translating natural language descriptions into complete audio productions. Users can specify styles, moods, instruments, and scenes in their prompts, and Fun-Music handles the rest: writing the lyrics, composing the melody, arranging the instrumentation, and rendering the vocals.
Two model variants are available: fun-music-v1 and fun-music-preview. Both include options for selecting male or female vocals, giving creators a baseline level of control over the output’s character.
API access follows standard authentication procedures, which means developers can integrate Fun-Music into their own applications and workflows rather than relying solely on a web interface.
Building on open-source roots #
Fun-Music doesn’t exist in a vacuum. Alibaba’s research division previously released InspireMusic, an audio generation project built on the FunAudioLLM framework, which was open-sourced in early 2025. That earlier project laid the groundwork for what Fun-Music now offers in a more polished, commercially oriented package.
Alibaba’s ATH division has also reportedly been working on an international AI music application called HappyShrimp. There’s no confirmed release date or public connection to Fun-Music at this point.
The timing aligns with Alibaba’s broader push into generative AI across multiple modalities. The company has been investing heavily in video generation capabilities alongside its text and audio work.
The competitive landscape is getting loud #
Alibaba is far from the first company to take a swing at AI music generation. Suno and Udio have both attracted significant attention in Western markets for their ability to create surprisingly listenable tracks from simple prompts. Google’s DeepMind has MusicLM, and Meta released MusicGen as an open-source offering. ByteDance, Alibaba’s domestic rival, has been developing its own audio AI capabilities as well.
What sets Fun-Music apart, at least on paper, is the end-to-end nature of the generation. Some competing models handle melody or arrangement but rely on external tools for lyrics or vocal synthesis. Fun-Music claims to handle the entire chain in a single pipeline, from the written word to the finished track.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our