StemTube Desktop 2.0.0 — Linux engines (installer superseded by 2.1.0)

⚠️ The installer here is superseded

Get the installer from linux-v2.1.0 instead.

The 2.0.0 .deb below ran the engine from a throwaway extraction on every
launch, so the in-app updater could never keep its changes: a 2.0.0 install
never actually received an update. The 2.1.0 installer fixes that.

The AppImage engines on this page are still current — they hold every
recent fix, and the 2.1.0 installer downloads them from here. This release is
kept for that reason; only its .deb is obsolete.

---

StemTube Desktop for Linux — self-contained AppImage

Two self-contained AppImages (mirror of the Windows model). Each bundles a
relocatable Python 3.12 + PyTorch + Demucs + madmom + Whisper + FFmpeg — no
Python, no pip, no system CUDA needed. Standard edition: local files only.

Which one?

• CPU — StemTube-x86_64-cpu.AppImage (~560 MB). Works on any 64-bit Linux.
• GPU — StemTube-x86_64-gpu.AppImage (~3 GB, split in 2 parts, CUDA 12.4).
  Needs an NVIDIA GPU + recent driver. Faster stem extraction.

Run it

chmod +x StemTube-x86_64-cpu.AppImage
./StemTube-x86_64-cpu.AppImage

Then the app opens in your browser. (Headless: add --no-window and open http://127.0.0.1:5011)

Requirements

64-bit Linux, glibc ≥ 2.35 (Ubuntu 22.04+ / Debian 12+ / recent Fedora, Arch…).
AVX-capable CPU (PyTorch needs it — will not run in a VM that masks AVX).

Prefer a one-command install that clones + pip-installs instead? A script is
also available (needs Python 3.10–3.13).