In-painting

AI-driven image restoration that removes unwanted objects and reconstructs missing pixels with seamless textures.

In-painting utilizes deep learning architectures (like LaMa or Stable Diffusion) to fill masked image regions by analyzing surrounding spatial context. It replaces manual cloning with automated synthesis: removing power lines from landscapes, erasing watermarks, or restoring damaged historical photos. By leveraging Fast Fourier Transforms (FFTs) and generative adversarial networks, the tech maintains high-resolution global consistency even in complex textures. It is the core engine behind Adobe Photoshop’s Generative Fill and Google’s Magic Eraser.

https://github.com/advimman/lama