Supported Browsers
WebGPU is the key requirement for local AI features. The table below shows which browsers support it and any flags or minimum versions you need:Not sure if your GPU is recognized? Open
chrome://gpu in Chrome or edge://gpu in Edge and look for “WebGPU: Hardware accelerated”. If it shows “Software only” or is unavailable, try enabling hardware acceleration in your browser settings or updating your GPU drivers.Hardware
Origami AI can run on modest hardware, but a dedicated GPU with F16 support makes a significant difference for local AI inference and video rendering speed:
A fast NVMe SSD is also recommended at the higher tier — model weights and project cache can be several gigabytes in size.
Local AI Models
When you enable local AI, Origami AI downloads the selected model to your browser’s cache on first use. Each model has a different footprint and capability focus:Models are cached in browser storage after the first download, so subsequent launches load instantly — even offline. If you’re on a device with limited VRAM, start with Llama 3.2 1B or Gemma 2 2B. You can always switch models in ⚙️ Settings → WebLLM without re-downloading others.
Node.js (Self-Hosting Only)
You only need Node.js if you plan to run Origami AI on your own machine rather than using the hosted app:
Check your version with
node -v. If you need to upgrade, use nvm or the official Node.js installer.
The hosted app at origami.techmitten.com requires no installation whatsoever — just a supported browser. Node.js is only needed when running the project locally from source or via Docker.