IOPaint (repository name Sanster/IOPaint, formerly lama-cleaner) is a free, open-source, fully self-hosted image inpainting and outpainting tool, powered by SOTA AI models.
Uses
- Remove unwanted objects, blemishes, watermarks, or people from images.
- Erase and replace any content in an image (powered by diffusion models such as Stable Diffusion).
- Perform image outpainting tasks.
Main capabilities
- Supports multiple AI models for erasing, inpainting, or outpainting tasks.
- Erase models: used to remove unwanted objects, blemishes, watermarks, or people from images.
- Diffusion models: used to replace objects or perform outpainting. Commonly listed models in the material include: runwayml/stable-diffusion-inpainting, diffusers/stable-diffusion-xl-1.0-inpainting-0.1, andregn/Realistic_Vision_V3.0-inpainting, Lykon/dreamshaper-8-inpainting, Sanster/anything-4.0-inpainting, BrushNet, PowerPaintV2, Sanster/AnyText, Fantasy-Studio/Paint-by-Example.
- Plugins: Segment Anything (accurate and fast interactive object segmentation), RemoveBG (remove image background or generate a mask for foreground objects), Anime Segmentation (similar to RemoveBG, specifically trained on anime images), RealESRGAN (super-resolution), GFPGAN (face restoration), RestoreFormer (face restoration), FileManager (conveniently browse images and save directly to the output directory).
How to use
- Provides a convenient WebUI for editing images with the latest AI models.
- IOPaint can be installed and started via the command line; after startup, all models will be downloaded automatically. To change the download directory, you can add the --model-dir parameter.
- After startup, you can visit in your browser http://localhost:8080 to get started.
- You can specify the plugins to use when starting the service; use iopaint start --help to see the commands for enabling plugins.
- Supports batch image processing via the command line: --image is the folder containing input images, --mask is the folder containing corresponding mask images; when --mask is the path to a single mask file, all images will be processed using that mask.
- Supports local sd ckpt/safetensors files (see the official documentation for specific usage).
- Frontend development: install nodejs, install frontend dependencies, create a .env.local file in web_app and fill in the backend IP and port, and start the frontend development environment; install backend dependencies and start the backend service, then visit http://localhost:5173/ for development. After modifying frontend code, it updates automatically; after modifying backend Python code, the service must be restarted.
Applicable scenarios
- Image object removal, watermark removal, blemish removal, people removal.
- Image content replacement and image outpainting.
- Image editing scenarios that require local self-hosting, emphasize privacy, or wish to use it completely free of charge.
- Supports CPU, GPU, and Apple Silicon.
- Windows provides a one-click installer; there is also OptiClean: an object eraser app for macOS and iOS.
Limitations
- The material does not provide specific minimum system requirements, model sizes, processing speed, charges, or pricing information.
- The material does not provide details on other features beyond the models and plugins mentioned above.
- Some models and plugins may require additional downloads or configuration; refer to the official documentation for specifics.
