Darktable

darktable is a photography workflow application and non-destructive raw developer, described by its own authors as a virtual lighttable and darkroom. It manages digital negatives in a database, shows them through a zoomable lighttable, and lets you develop, enhance and export them to local or remote storage.

Editing never rewrites the original file, and the project states plainly that it is not a free Adobe Lightroom replacement, worth knowing before switching.

Platform and hardware expectations are set out unusually clearly for a project of this size.

• Supported systems: Linux, FreeBSD, NetBSD, OpenBSD, Windows 10 or later, and Apple Silicon Macs on macOS 14 or later.
• Recommended hardware: 8 GB of RAM, a quad-core Intel Core i5, and an OpenCL 1.2 capable GPU.
• GPU guidance: a GPU is optional but strongly recommended, and Nvidia cards are the safer bet.
• Optional AI features: object masks, denoise and upscale, compiled in only with a build flag and off by default.
• Scripting: a Lua API with its own reference documentation, maintained in a separate repository.

CPU inference for the AI features is bundled and works out of the box, with CoreML on Apple Silicon and DirectML on Windows included. GPU acceleration through CUDA, ROCm or OpenVINO requires a separately installed build of ONNX Runtime.

Linux is where darktable is developed, and the team recommends it wherever there is a choice. It runs on modest machines, even a Raspberry Pi, though modules such as denoise, local contrast and retouch turn slow enough to be unusable.