Overview
Open Generative AI is an open-source (MIT) studio app for generating images, videos, audio and lip-synced talking clips. It ships as an Electron desktop app for macOS, Windows and Linux, and as a Next.js web app you can run yourself. The interface is split into studios: Image, Video, Audio, Lip Sync, Cinema, Workflow, Agent and several more, each with a prompt box, a model picker and controls that adapt to the chosen model.
It is important to understand where the models run. Most of the catalogue, which the README counts at 400+ model definitions including Flux, Nano Banana, Midjourney, Kling, Sora, Veo, Seedance and Wan, is not run by the app itself: requests are sent to the Muapi.ai API (submit a job, then poll for the result) using a Muapi access key you paste in on first use. The code is open and self-hostable, but hosted generation depends on that paid third-party gateway. The same studio component library also powers the hosted version on muapi.ai.
The desktop app adds a genuinely local path for some models. A bundled stable-diffusion.cpp engine runs image models such as SD 1.5, SDXL and Z-Image on the machine itself (Metal on Apple Silicon, CUDA/Vulkan/ROCm elsewhere) with no API key, and the app can connect to a Wan2GP server you run on your own GPU box for Flux, Qwen-Image and video models like Wan 2.2, Hunyuan Video and LTX Video. Local inference is only available in the desktop app; the web version always uses cloud APIs.
What it does
- Image Studio and Video Studio that switch automatically between text-to-X and image-to-X model sets depending on whether you upload a reference image or start frame.
- Lip Sync Studio with two modes: a portrait image plus audio becomes a talking video, or an existing video plus audio becomes a lip-synced video, across 9 dedicated models.
- Multi-image input of up to 14 reference images for compatible edit models, with an ordered multi-select picker and a local upload history.
- Workflow Studio: a node-based editor for chaining image, video and audio models into multi-step pipelines, with templates and a playground to run them.
- Optional local inference in the desktop app via a bundled sd.cpp engine (SD 1.5, SDXL, Z-Image) or a user-run Wan2GP server for Flux, Qwen-Image and video models.
- Cinema Studio camera controls (camera body, lens, focal length, aperture) that are translated into prompt modifiers for cinematic shots.
Getting started
Most users can download a prebuilt installer for macOS, Windows or Linux from the GitHub Releases page; the installers are not code-signed, so macOS Gatekeeper and Windows SmartScreen will warn on first launch. The steps below follow the README's build-from-source path. Hosted models need a Muapi.ai access key; local models in the desktop app do not.
Check the prerequisites
You need Node.js v18 or newer, plus a Muapi.ai access key (copy the key value itself, not its name) if you want to use hosted models.
Clone the repository with submodules
The workflow and agent packages live in git submodules, so clone recursively.
git clone --recurse-submodules https://github.com/Anil-matcha/Open-Generative-AI.git
cd Open-Generative-AIInstall and build the workspaces
The README notes that npm install alone is not enough: this script installs dependencies and builds the studio, workflow and agent packages.
npm run setupStart the desktop or web app
Run the Electron desktop app (recommended) or the Next.js web version at http://localhost:3000. Enter your Muapi API key when prompted, or skip it if you only plan to use local models.
npm run electron:dev # desktop app (Electron + Vite)
npm run dev # web version (Next.js)Optional: enable local image models
In the desktop app, open Settings → Local Models, install the sd.cpp inference engine with one click, download a model, then flip the ⚡ Local toggle next to the model selector in Image Studio. To use video models locally, run a Wan2GP server on a CUDA or ROCm GPU machine and paste its URL under Settings → Local Models → Wan2GP server.
# On your GPU machine
git clone https://github.com/deepbeepmeep/Wan2GP
cd Wan2GP
./install.sh
python wgp.py --listen --server-name 0.0.0.0Commands and code are distilled from the project's own documentation — always check the official repo for the latest.
When to use it
- Try many commercial and open image and video models from one interface with a single Muapi.ai key instead of separate accounts per provider.
- Turn a portrait photo and a voice track into a talking-head video, or re-sync the lips on an existing clip.
- Generate SD 1.5, SDXL or Z-Image images fully on your own Mac or PC with the desktop app's bundled sd.cpp engine, without an API key.
- Fork and rebrand a ready-made generation studio UI as the front end for your own creative product.
How Open Generative AI compares
Open Generative AI alongside other open-source video generation tools AI/TLDR tracks, ranked by GitHub stars.
| Tool | Stars | What it does |
|---|---|---|
| MoneyPrinterTurbo | ★ 134k | A self-hosted WebUI, API and CLI that turns one topic or keyword into a finished short video — script, stock or generated footage, voiceover, subtitles and music. |
| OpenMontage | ★ 70.9k | An agentic video production system that gives a coding assistant 12 pipelines and 100+ tools to research, script, generate and render a video end to end. |
| HyperFrames | ★ 70k | HeyGen's HTML-to-video renderer: author a composition in HTML, CSS and JS, preview it in a browser, then render the same input to a deterministic MP4 locally or on AWS Lambda. |
| Open Generative AI | ★ 31.3k | MIT-licensed image, video and lip-sync studio app that runs hosted models through the Muapi.ai API |
| Hypit | ★ 24.8k | A markup language and compiler that lets a coding agent rebuild a reference video as an editable workflow — footage, captions, B-roll and effects anchored to words rather than timestamps — and render variants in one command. |
| Toonflow | ★ 19k | Infinite-canvas workbench for AI short-drama production, taking a novel through chapter events, script, storyboards and video nodes. |
| waoowaoo | ★ 14.7k | Self-hosted creative workspace for AI image and video: a canvas holding references and generated assets, a side Assistant that develops the brief, and model-aware controls over your own OpenRouter key. |
| Palmier Pro | ★ 14.6k | A Swift-native macOS video editor with generative models in the timeline and a built-in MCP server, so Claude, Codex or Cursor can edit the open project alongside you. |