🤖 Our AI models
Every AI model in our tools is open-licensed for commercial use and runs in your browser — here are the details.
Some of our tools use AI models (background removal, image upscaling, photo-to-line-art). All of them are open source under permissive licenses (Apache-2.0, BSD, MIT) — free to use, including commercially.
🔒 Important: all AI processing runs in your browser (client-side) — your images are never sent to any server. Whatever you create is entirely yours and free for commercial use, with no attribution required.
BiRefNet (lite)
MIT· ~224MB‘Ultra’ background removal — a state-of-the-art network that produces the sharpest edges (hair, fur, fine detail), well beyond U²-Net. Powers the remove-background tool's Ultra quality.
U²-Net
Apache-2.0· ~168MBBackground removal — a segmentation model that finds the subject and separates it from the background. Powers the remove-background tool (Best & Fast quality). Redistributed via rembg (MIT).
Real-ESRGAN ×4
BSD-3-Clause· ~67MBAI upscaling (super-resolution) — enlarges an image 4× without losing sharpness. Used in the DPI tool to upscale images for engraving.
Informative Drawings
MIT· ~17MBPhoto → line art — a neural network that produces clean outlines from a photo (Caroline Chan et al., CVPR 2022). The same weights used by ControlNet's lineart preprocessor. Powers the line-art tool.
This page is provided for transparency and is not legal advice. The binding licenses are those of the upstream projects linked above.