Local WebGPU
The image and model inference stay on this device. Only static site code and the versioned model are downloaded.
Free browser tool
Upscale an image 2× with local AI.
Use your device’s GPU to create a larger image without uploading it. This bounded release accepts one opaque JPG, PNG, or WebP up to 2 MB, with a browser-specific pixel ceiling shown after the worker check.
Dedicated GPU workerNo image upload2 MB maximum
AI Image Upscaler
Checking WebGPU support…
Quick feedback
Did this result work for you?
One tap is enough. You can ignore this and keep working.
Thanks — that helps us improve.
Bounded first release
One opaque image, 2× only, up to 2 MB. Verified Windows desktop browsers allow 307,200 input pixels; Safari receives a smaller 76,800 pixel profile until physical-device evidence supports more.
Generated detail
Upscaling can make edges and texture look clearer, but generated details are not guaranteed to match the original scene.
Why is my browser unsupported?
This release requires WebGPU inside a dedicated worker, a working adapter/device, and a successful model session. Chrome, Edge, and Firefox on Windows have repeated browser evidence. Safari is kept on a smaller profile while Mac, iPhone, and iPad evidence is collected. Resize Image remains available without AI.
Does BeanNest upload my photo?
No. The source image, decoded pixels, AI output, and download remain in browser memory. BeanNest analytics never receive image bytes, filenames, thumbnails, or exact dimensions.
Why reject transparent images?
The selected model produces RGB pixels. This first release rejects alpha instead of silently flattening it against an unexpected background.