A Linux distro for local AI, with the Omni Studio desktop workspace.
Important
Portable Omni Server is a Linux distro. Windows users must install WSL2 and enable hardware virtualization. The desktop window also requires Microsoft WebView2. CPU-capable workloads can run without an NVIDIA GPU; CUDA workloads require a compatible NVIDIA GPU and Windows driver with WSL support. Python is included in the portable release.
Maintained by aivrar.
The distro and Windows launcher were built from my portable-linux-in-a-box project. Omni Studio adds the local AI workspace, gateway and model tools.
Omni Studio is a local multimodal workspace packaged as a Windows application with a dedicated WSL2 Linux distro. The Windows window talks to a local bridge and gateway; models, workflows, and generated media live inside that distro.
The live Home page in its idle setup state. Browse the screenshot gallery for every workspace.
- Chat: local model sessions with model-specific text, image and audio input.
- ComfyUI: workflow import, requirements checks, GPU placement and queue controls.
- Music and audio: ACE-Step, Audio Lab and Music 3 workspaces, plus MOSS sound effects.
- Media library: browse, play, organize and export saved results.
- Model library and Runtime: install models, select devices and manage workers.
- CLI and API: automate the same local studio from scripts and other clients.
Models, workflows and generated media are stored inside the Linux distro. Choose the engines and weights needed for your work. See feature compatibility for supported modalities, engine requirements and task-specific behavior.
| Requirement | Purpose |
|---|---|
| 64-bit Windows 11, or Windows 10 21H2 or later | Windows host for the launcher and WSL GPU workloads |
| WSL2 and hardware virtualization | Runs the Linux distro |
| Microsoft WebView2 Runtime | Displays the desktop workspace |
| Free disk space | Holds the growing Linux disk, model downloads and generated media |
| Internet access for downloads and updates | Obtains the release image, model weights and optional components |
CPU or GPU: the workspace and CPU-capable workflows can run on CPU. CPU generation can be much slower and needs enough system RAM. A compatible NVIDIA GPU and Windows driver enable CUDA acceleration; some models and custom nodes specifically require CUDA. Check the selected engine's compatibility requirements.
Python is already included. The Linux image contains Python and the model
environments. The Windows package also contains an embedded Python under
runtime/python/ for its networking relay. Neither requires a separate Python
installation on the user's PC.
Install WSL from PowerShell as Administrator, then restart Windows:
wsl --installAfter restarting, check wsl --status. See Microsoft's
WSL installation guide
and GPU setup guide.
Omni Studio uses its own distro named linbox-Omni_Studio.
Download the Windows package from Releases.
Extract it, run Prepare-Omni.cmd to download and verify the preinstalled Linux
image, then launch Omni_Studio.exe. Keep the app folder together. Model weights
are selected and downloaded inside the app. Follow
Launch and first run.
The GitHub source checkout contains the application code and documentation;
the Windows launcher binaries, Linux runtime image and model weights are
separate package components. See package contents.
- GitHub wiki: the full illustrated operator manual with page navigation.
- Usage manual: launching the packaged app, each UI tab, the CLI, GPU placement, shutdown, and feature compatibility.
- API capability inventory: route families
and authentication. The typed request models in
server/routers/are the exact request contract. - Feature compatibility: model inputs, engine-specific requirements and supported tasks.
- Repository layout: source and local runtime locations.
- Development guide: source setup and focused checks.
| Path | Role |
|---|---|
server/ |
Gateway, workers, setup, and browser UI |
cli/, omni-cli, omni-cli.bat |
API command-line client and launchers |
comfy_nodes/ |
Omni-owned ComfyUI extension |
manual/, docs/, skills/ |
Operator, API, and agent guidance |
tests/, test_assets/ |
Regression tests and reusable fixtures |
bridge.py, bridge_watchdog.py, windows_loopback_relay.py, app.json |
Local launcher contracts |
reports/ |
Dated investigation records, including the repository audit |
Original project code is available under the MIT license. See
third-party and media terms. Unreviewed binary fixture
and report media remains local and is excluded from the default Git release.
Reviewed app screenshots are included under manual/images/; their captions,
capture details, and asset hashes are linked from the screenshot gallery.