Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/anil-matcha/open-generative-ai/llms.txt

Use this file to discover all available pages before exploring further.

This guide walks you through cloning the Open Generative AI repository, running the setup script, and generating your first AI image or video. By the end you’ll have a fully working local studio — either as a desktop app powered by Electron or as a web app running on http://localhost:3000 — connected to 200+ models via your Muapi API key.

​
Prerequisites

Before you begin, make sure you have the following:
  • Node.js v18 or later — nodejs.org
  • A Muapi.ai access key — generate one at muapi.ai/access-keys. Copy the generated key value; do not use the key name or label.
Most users should download the desktop app instead of building from source — no Node.js or terminal required. See the Deployment page for one-click installers for macOS, Windows, and Linux.

​
Steps

1

Clone the repository with submodules

Open Generative AI uses Git submodules for the workflow builder, agent, and design-agent packages. Cloning without --recurse-submodules will leave those packages empty and break the build.
git clone --recurse-submodules https://github.com/Anil-matcha/Open-Generative-AI.git
cd Open-Generative-AI
If you already cloned without the flag, initialize the submodules now:
git submodule update --init --recursive
2

Run setup

The setup script does three things in one command: initializes any remaining submodules, installs all npm dependencies across every workspace, and builds the workspace packages (studio, workflow-builder, agents, design-agent). Running plain npm install is not sufficient — the workspace packages must be compiled before either dev script will work.
npm run setup
This may take a few minutes on the first run. You’ll see each workspace package compile in sequence.
3

Start the app

Choose the mode that fits your workflow:Desktop app (recommended) — launches the Electron wrapper backed by a Vite build:
npm run electron:dev
Web app — starts the Next.js dev server at http://localhost:3000:
npm run dev
The desktop app is the only mode that supports local inference via the bundled sd.cpp engine and the Wan2GP remote server. If you plan to generate images or videos without an API key, use npm run electron:dev.
4

Enter your Muapi API key

On first launch, the app displays an API key prompt. Paste the key value you copied from muapi.ai/access-keys and confirm. The key is stored in localStorage and is sent only to the Muapi API — never to any other server.You can update or remove your key at any time from the settings panel inside the app.
5

Generate your first image or video

With the app running and your API key saved:
  1. Open the Image Studio to generate images, or the Video Studio to generate videos.
  2. Choose a model from the model selector. Start with something fast — Flux Dev for images or Kling v1.5 for videos work well as first tests.
  3. Type a prompt in the prompt field.
  4. Click Generate.
Results appear in the generation panel and are saved to your local generation history automatically. Click any result to download it at full resolution.

​
Troubleshooting

This error means Next.js can’t locate the app/ directory. It almost always has one of two causes:
  1. You’re not running the command from the repo root. Make sure your current directory contains app/, package.json, and next.config.mjs before running npm run dev.
  2. Submodules were not cloned. If packages/Vibe-Workflow or packages/Open-Poe-AI are empty directories, the workspace build will fail silently. Re-run the full setup: 
git submodule update --init --recursive
npm run setup
If npm run setup exits with errors, try running each step individually to isolate the problem:
# Step 1 — ensure submodules are present
git submodule update --init --recursive

# Step 2 — install all dependencies
npm install

# Step 3 — build workspace packages in dependency order
npm run build:packages
Check that each packages/ subdirectory contains source files before running build:packages.

​
Next Steps

Image Studio

Explore 50+ text-to-image and 55+ image-to-image models, including multi-image input support.

Video Studio

Generate videos from text prompts or animate a start-frame image with 100+ models.

Local Inference

Run image generation entirely offline with the bundled sd.cpp engine on Mac, Windows, or Linux.

Build docs developers (and LLMs) love

Get started for freeTalk to us