Pymss vs StemRoller
Pymss
pymss.netStemRoller
stemroller.comSide by side
| Figure | Pymss | StemRoller |
|---|---|---|
| Price from | Free | Free |
| Free plan | Yes | Yes |
| Free trial | — | — |
| Open source | Yes | — |
| Web | Yes | — |
| Windows | Yes | Yes |
| macOS | Yes | Yes |
| Linux | Yes | — |
| iOS | Yes | — |
| Android | Yes | — |
| API | Yes | — |
| Self-hosted | Yes | — |
| Plans published | 1 | 1 |
| Spec score · AI stem splitters | 8.7 (#5) | 7.6 (#20) |
| Integrations | The project provides ComfyUI custom nodes and a VST3 ARA plugin for use in DAWs. | StemRoller uses YouTube search results to find songs, according to its GitHub README. |
| Platforms | The official site lists Windows, macOS, and Linux for Studio and CLI, iOS and Android for the mobile runtime, and Windows for the ARA plugin. | The maker's releases page provides downloads for Windows and macOS. |
| Purpose | Pymss is an open source music source separation toolchain that separates vocals, drums, bass, and other stems on the user's machine. | StemRoller creates karaoke, vocal, and stem tracks from songs. |
| Support | The project points users to GitHub documentation and an issue tracker, and its Studio README lists a QQ group for Chinese beta testing, bug reports, and community discussion. | The maker offers help through its Discord server. |
Plans
In their own words
Pymss
Pymss is an open-source music source-separation toolchain that runs on the user's own machine and separates audio into vocals, drums, bass, and other stems. Its Studio desktop workspace handles model management, batch jobs, and stem editing. Users can import files or folders, choose a downloaded model, configure inference, and queue work. Studio can browse, download, resume, and delete model files, with storage cleanup controls. Waveform previews, mix controls, saved projects, and stem export support work on the results. Pymss can chain models so each workflow step uses the preceding step's output. The inference package lists WAV, FLAC, MP3, M4A, AAC, Opus, Vorbis, and Ogg output. Other components include ComfyUI custom nodes, a VST3 ARA plugin for compatible DAWs, and a browser interface bundled with the server. The ARA plugin provides per-stem mute and solo controls; its README gives Reaper and Studio One as host examples. The repositories have separate licenses: Studio is listed under AGPL-3.0 and the inference package under MIT. Studio remains at an early development stage, and changes between versions may break things.
StemRoller
StemRoller is a free desktop application for turning songs into karaoke, vocal, and instrumental material. Search by song title and it finds results online through YouTube search integration. Using the Demucs machine-learning project, StemRoller separates a song into four outputs: vocals, drums, bass, and everything else. It also mixes the non-vocal stems into an instrumental track. Splitting takes several minutes, and batch processing is not available. The maker describes StemRoller as open source, with the README offering either the Unlicense or MIT No Attribution license. The homepage links to a macOS ZIP and a Windows CUDA executable, version 3.1.1. Linux is not officially supported, although the README provides manual dependency setup instructions. The Free plan costs nothing and includes four stem outputs plus an instrumental track. Questions and issues can be taken to Discord.