SongGeneration Studio is a clean, polished open-source interface for Tencent AI Lab’s SongGeneration model (LeVo), built to create complete AI songs with vocals, lyrics and full instrumental tracks. You can start from a text prompt and lyrics, or from reference audio, and produce a finished song locally. The project has about 606 stars on GitHub and was updated in January 2026.
Running a research music model from the command line is fiddly; SongGeneration Studio wraps it in a friendly desktop-style interface distributed through Pinokio for one-click installation. Creators control the style in detail: genre from pop and rock to jazz and folk, mood, male or female voice with customizable timbre, and the instrument palette. Batch processing lets you queue multiple generations, and smart model selection helps pick the right model variant for the job. Because generation runs on your own GPU, the project notes a minimum of about 10 GB of VRAM, which keeps songs private and avoids per-generation fees. Under the hood it is Tencent’s SongGeneration doing the singing; the studio’s contribution is making that model genuinely usable.
The studio targets users who find research code releases hard to operate: installation, model selection and generation controls are presented as ordinary application workflows rather than command-line flags and configuration files, which lowers the barrier for non-programmers.
SongGeneration Studio is free to download from GitHub. Generation runs locally on your own hardware using Tencent’s SongGeneration model, so there are no per-song charges.
SongGeneration Studio is for musicians, producers and hobbyists who want to experiment with full AI-generated songs on their own machine rather than through a hosted music service.