BROWSER-BASED IMAGE INPAINTING
Select a visible watermark on one image and use Apply to all to reuse that selection on a batch with matching relative positioning. The editor supports up to 30 PNG, JPG/JPEG, or WebP images per session; practical performance depends on your device and browser.
Only edit images you own or have permission to modify. Removing a visible watermark does not grant usage rights or remove invisible provenance markers.
REPEAT A MANUAL SELECTION
EraseKit applies one manually drawn mask across images. It does not automatically detect marks or guarantee that every image in the batch has the same watermark.
Load up to 30 PNG, JPG/JPEG, or WebP images. Similar crop and framing make shared coordinates more likely to line up.
Paint over the full visible watermark on one image. Keep the mask as tight as practical while covering its edges.
Use Apply to all when the mark occupies the same relative position in every image. Review each result; edit images one by one if placement differs.
The editor stores the selection proportionally to the source image dimensions and scales it for the other images. A logo a similar distance from the lower-right corner may align across matching exports. It will not track a moving, resized, rotated, or repositioned watermark. Mixed orientations and crops can put the mask over clean content.
The inpainting model initializes when editing begins and may need to download first. ONNX Runtime Web uses WebGPU where available and WebAssembly when needed. Image processing runs in the browser; site and model downloads still use the network.
Accepted inputs are PNG, JPG/JPEG, and WebP. Downloads retain the corresponding format; JPEG and WebP are re-encoded. Canvas dimensions are retained, but large selected regions may be processed smaller and scaled back, which can soften detail.
Although the interface accepts up to 30 images per session, processing is sequential. Time and feasibility vary with image dimensions, selection size, available device memory, graphics support, and browser capabilities. A large batch can be slow or fail on constrained devices. Try a smaller set if the browser struggles.
No. You manually select one region and choose whether to reuse its coordinates across the batch.
The interface accepts up to 30 images per session. This is a practical interface cap, not a guarantee that every device can process a batch of that size.
Process each image separately so you can paint a mask at its actual position.
The inpainting operation runs in your browser and images are not sent to an EraseKit processing server. The site and model resources do load over the network.
Only if you have permission to edit and use that image. Do not remove marks from stock previews, proofs, or unlicensed assets.
