PowerPaint V2

A Task is Worth One Word: Learning with Task Prompts for High-Quality Versatile Image Inpainting

GitHub: https://github.com/open-mmlab/PowerPaint (opens in a new tab)

Object insertionObject RemovalShape-guided Object InsertionOutpainting
Original Imagecropinputcropinputimagecropinput

PowerPaint introduces the concept of "learnable task prompts" to guide the model in achieving specific tasks more effectively. These tasks include text-guided, shape-guided, object-remove, and outpainting.

In PowerPaintV2, the author used the same method as BrushNet to train a model that can transform any sd1.5 base model into an inpainting model. When you use any SD1.5 base model(e.g: runwayml/stable-diffusion-v1-5), the option for PowerpaintV2 will appear in the sidebar.


You can select specific tasks through the dropdown menu on the right. The outpainting task is automatically utilized when using Expander.


Some configurations (such as ControlNet, LCM LoRA...) cannot work with PowerPaintV2, and they will be automatically disabled when you enable PowerPaintV2.