Open-Generative-AI: the open-source AI image & video studio
Open-Generative-AI is a free, open-source (MIT) studio on GitHub for image, video, lip sync and music generation, with 600+ models listed. Some models run locally on your own GPU with no API credits; the cloud models need a paid API key. Here is how to install it and what is really free.
What it is (and what's free)
- The repo: github.com/Anil-matcha/Open-Generative-AI, about 34k stars, MIT license, self-hosted.
- Local models (free to run): image models through sd.cpp, which comes with the app and works on a Mac with Apple Silicon (for example Z-Image, SDXL). Video and large image models (Wan 2.1/2.2, Hunyuan Video 1.5, LTX-2, Flux.1 Dev, Qwen Image) go through Wan2GP, a separate tool you install yourself on a machine with an NVIDIA or AMD graphics card. No API key, no credits.
- Cloud models (paid): the big hosted models (Kling, Veo, Sora, Midjourney…) go through a Muapi API key, billed by that provider. You can skip the key if you only use local models.
Local video does not run on a Mac alone: it needs Wan2GP on an NVIDIA or AMD graphics card with enough VRAM. Check the README of the model you pick before downloading it: some are tens of GB. Star counts and model lists change; check the repo itself today.
Install: the easy way
- Open the repo's Releases page and download the installer for your system: .dmg (Mac), .exe (Windows) or .AppImage (Linux).
- Install and open it like any app.
- When it asks for an API key, skip it if you want local models only.
Install: from the code (if you prefer)
You need Git and Node.js installed. Then, in a terminal:
git clone --recurse-submodules https://github.com/Anil-matcha/Open-Generative-AI.git cd Open-Generative-AI npm run setup npm run electron:dev # desktop app # or: npm run dev # in the browser
If a command fails, read the README of the repo: it is the source of truth and may have changed since this guide.
Your first local generation
- Pick a local image model first (smaller download, faster), not a video model.
- Let it download, write a short prompt, generate.
- Once image works, try a local video model. Expect long generation times on a normal laptop.
Frequently asked questions
Is it really free?
The app is free and open source. Local models cost nothing per generation, only your computer's power and disk space. Cloud models are paid through the API key.
"No filters": what does that mean for me?
The project describes itself as having no content filters. That doesn't change the law or platform rules: you stay responsible for what you generate and post (people's likeness, copyright, explicit content).
My computer has no GPU. Can I use it?
Local video will be very slow or won't run. You can still use the cloud models with a paid key, or stick to small image models.
Want the next one before everyone else?
One short email when a new free guide is out. Nothing else.
Rather not pay per generation?
Our own reels are not generated: Claude writes each one as a web page and we export it to MP4. The first three lessons of the method are free to read.