chaiNNer builds visual pipelines for AI image upscaling and batch processing.

chaiNNer is not a one-click upscaler: its node-based workflow requires you to assemble a processing chain, and beginners may need time to understand how the nodes fit together. Processing speed also depends on having a reasonably modern, capable GPU, so the app is less suited to machines without suitable graphics hardware.

Once a pipeline is set up, chaiNNer can handle a single image or apply the same sequence of operations to a folder. You place processing nodes on a canvas and connect them to define how images are loaded, transformed, and saved. This visual approach supports repeatable work without requiring you to write processing scripts.

For example, a chain can connect image loading, an upscale or restoration operation, and image saving, while color-coded handles indicate compatible connections. In chaiNNer, model processing can use PyTorch, NCNN, ONNX, or TensorRT, giving users several framework options for building image workflows.

Preparing the processing environment

The program manages its own isolated Python environment, so a separate manual Python setup is not required. On first launch, it downloads the integrated Python installation. Its Dependency Manager is then used to install the AI frameworks needed for processing; those framework downloads can be substantial, and the models themselves are obtained separately.

Before adding an AI model, install the matching framework through the Dependency Manager and make sure the model is available for the relevant nodes. chaiNNer does not include upscaling models, so selecting and loading a model is a separate part of preparing a pipeline. Keeping these steps distinct helps clarify whether a problem relates to the processing setup or to a missing model.

Choosing a framework and building a batch chain

The framework choice can depend on the available hardware. The developers recommend PyTorch or TensorRT for NVIDIA cards and NCNN for AMD. chaiNNer also supports ONNX, allowing workflows to use any of the listed frameworks when appropriate. The notes do not claim that every combination delivers the same performance, so users should select a supported option suited to their system and model.

For batch work, connect the required operations once and apply that chain to a folder of images. This can make repeated upscaling or restoration more consistent than setting up each image independently. The same node-based method also supports more involved processing sequences, while keeping the order of operations visible on the canvas.

Platforms and project details

chaiNNer runs on Windows, macOS, and Linux. It is an open-source project released under the GNU GPL v3, with development credited to chaiNNer contributors. Its origins as an AI upscaling tool have expanded into a general node-based image processor, but its practical focus remains visual workflows for image operations, particularly AI upscaling, restoration, and batch processing.

chaiNNer is a practical fit for users who want to build and reuse image-processing pipelines rather than rely on a single automatic action. Its main trade-offs are the learning curve, the separate setup of frameworks and models, and the need for capable hardware to achieve good speed. Understanding those requirements before assembling a workflow makes it easier to plan both the model setup and the processing chain.