Repo Image & Video

Anil-matcha/Open-Generative-AI

Open-source AI image, video, lip sync and cinema studio powered by Muapi, offering desktop apps, a hosted version and optional local inference via sd.cpp or Wan2GP.

  • 29.9k GitHub stars
  • JavaScript
  • ⚖️ MIT
  • 🎯 Intermediate
View on GitHub (opens in a new tab) Setup guide
npm run electron:build:linux
Anil-matcha/Open-Generative-AI preview image

What it is

Open Generative AI is a free, open-source studio for AI image, video, cinema and lip sync generation. It is powered by Muapi.ai and covers text-to-image, image-to-image, text-to-video, image-to-video and audio-driven lip sync across many models. It is available as a hosted web version and as desktop installers for macOS, Windows and Linux. The desktop app can also run some models locally through sd.cpp or a user-installed Wan2GP.

Who it's for

  • Creators who want a self-hosted, customizable alternative to subscription AI image and video platforms
  • Users who want to run image models locally on a Mac, Windows or Linux desktop
  • Users with a CUDA/ROCm GPU machine who want local video generation through Wan2GP
  • Builders who want to extend the studio with their own models or UI changes

Requirements

Requirements

  • Desktop installers need no Node.js or terminal; a Muapi API key is used for cloud generation
  • Local sd.cpp inference is available only in the desktop app; 16 GB RAM is recommended for Z-Image models
  • Wan2GP local video needs a user-installed Wan2GP on a machine with a CUDA or ROCm GPU
  • libfuse2 may be needed to run the AppImage on older Linux systems

Setup

  1. Download the desktop app

    Download the installer for your platform from the GitHub releases page: macOS Apple Silicon (.dmg), macOS Intel (.dmg), Windows x64 (.exe), or Linux (.AppImage / .deb from the v1.0.9 release). Or use the hosted version in the browser with a free account.

  2. macOS: clear Gatekeeper block

    The app is not notarized by Apple. Drag the app to /Applications, run this command in Terminal, then right-click the app, choose Open, and click Open again.

    bash
    xattr -cr "/Applications/Open Generative AI.app"

  3. Windows: SmartScreen warning

    The installer is not code-signed. Click More info on the SmartScreen dialog, then click Run anyway. The app installs to %LocalAppData% with a Start Menu shortcut.

  4. Linux: build installers locally

    Build the AppImage and .deb with Electron Builder. Output is written to the release/ folder.

    bash
    npm run electron:build:linux

  5. Enable local sd.cpp inference

    In the desktop app, open Settings → Local Models, install the sd.cpp engine (one click), and download a model (plus auxiliary files for Z-Image). Then in Image Studio click the ⚡ Local toggle next to the model selector and pick your local model. No API key is needed for local generation.

  6. Optional: install Wan2GP for local video

    Install Wan2GP yourself on a machine with a CUDA or ROCm GPU, then set its folder in Settings → Local Models → Wan2GP, click Save, then Check.

    bash
    git clone https://github.com/deepbeepmeep/Wan2GP
cd Wan2GP
./install.sh                          # or install.bat on Windows

Examples

Run the AppImage on Linux

bash
bash
chmod +x "release/Open Generative AI-*.AppImage"
./release/Open\ Generative\ AI-*.AppImage

What it does: Makes the built AppImage executable and launches it after running the Linux build.

Start a Wan2GP MCP server for a remote machine

bash
bash
python wgp.py --mcp --mcp-api-version 1 --mcp-transport streamable-http --mcp-host 0.0.0.0 --mcp-port 7866

What it does: Lets the desktop app, for example on a Mac, offload local video inference to a GPU machine. The README says to use --mcp-host 0.0.0.0 only on a trusted network.

Verify sd.cpp directly with sd-cli

bash
bash
DYLD_LIBRARY_PATH="$APP_DATA/bin" "$APP_DATA/bin/sd-cli" \
  -m "$APP_DATA/models/DreamShaper_8_pruned.safetensors" \
  -p "a serene mountain lake at sunrise, oil painting" \
  -o /tmp/sd15-test.png \
  --steps 12 -H 512 -W 512 --cfg-scale 7.5 --seed 42 \
  --sampling-method euler_a

What it does: Runs a single 512x512, 12-step inference with the same binary the app uses, as a sanity test on Mac. The README sets APP_DATA and downloads the Dreamshaper 8 model in earlier steps.

Pros & cons

Pros

  • Pro:Many studios in one app: Image, Video, Audio, Lip Sync, Cinema, Workflow, Agent and more
  • Pro:Desktop app supports local inference through bundled sd.cpp or a user-installed Wan2GP, in addition to cloud models
  • Pro:Multi-image input accepts up to 14 reference images on compatible edit models such as Nano Banana 2 Edit
  • Pro:Generation history and upload history are stored locally for reuse across sessions

Cons

  • Con:Desktop installers are not notarized (macOS) or code-signed (Windows), so users must work around security warnings
  • Con:Local inference is desktop-only; the hosted web version always uses cloud APIs
  • Con:Wan2GP local video requires your own install and a CUDA/ROCm GPU, and Z-Image is known to hang a base 8 GB M-series Mac

Images

Anil-matcha/Open-Generative-AI preview image
Anil-matcha/Open-Generative-AI preview image · Anil-matcha/Open-Generative-AI on GitHub · MIT
Anil-matcha/Open-Generative-AI preview image