Image inpainting is the AI technique that fills in missing or unwanted areas of a picture by reconstructing them from the surrounding pixels. This free online tool puts two state-of-the-art models in your browser — LaMa and MI-GAN. Brush over a damaged scan, an unwanted object, a caption or a watermark and the network regenerates the area seamlessly. No upload: the models run locally on your device.
Open any JPG, PNG or WEBP with a damaged or unwanted area.
Brush over scratches, tears, text, logos, objects or people you want gone.
Pick LaMa for large areas or MI-GAN for speed — the AI reconstructs the missing pixels from the surroundings.
Save the repaired image at your original resolution.
Inpainting is the process of filling missing or masked pixels in an image. AI inpainting models like LaMa and MI-GAN analyze the surrounding pixels and generate a realistic fill for the covered area.
MI-GAN is a fast model ideal for small watermarks, text and blemishes. LaMa is an HD model that produces the best results on large masked areas such as removed objects or damaged sections of old photos.
Yes — scratches, creases, stains and torn corners can be painted over and reconstructed. For very large damage, use LaMa with a wide brush.
Yes — free, no account, no upload. The inpainting models run in your browser, so your image never leaves your device.