chaiNNer

chaiNNer is a desktop editor for people who want AI upscaling and repeatable image processing without writing scripts. Its visual workflows combine model processing with ordinary edits, so you can reuse the same chain across a collection of images.

• AI upscaling. Run community models from OpenModelDB, including ESRGAN, SPAN, and OmniSR architectures supported through Spandrel, with face restoration and corrections for color shifts caused by upscaling.
• Image editing. Combine background removal, color adjustments, sharpening, blur, resizing, cropping, blending, and format conversion in one workflow.
• Textures and pixel art. Generate and combine normal maps, convert PBR textures, export DDS files with mipmaps, and process spritesheets; pixel art tools include non-AI scaling, palette extraction, and dithering.
• Batch processing. Apply workflows to image folders or video frames, compare models, and skip upscaling when an image is already large enough. Saved workflows can also run from the command line.

chaiNNer is free and open source under GPL-3.0, with installers and portable builds for Windows, macOS, and Linux. Models run locally through PyTorch, NCNN, ONNX, or TensorRT. It supports NVIDIA, AMD, Intel, and Apple Silicon GPUs depending on the backend, with CPU processing as a fallback. AMD's PyTorch acceleration requires Linux; Apple Silicon supports PyTorch MPS. First launch needs an internet connection to download its Python environment. Windows 8.1 and earlier and macOS 10.x and earlier aren't supported.