Rembrandt Offers Free Photo Editing That Stays on Your Machine

Rembrandt is a free photo editor for macOS, Windows and Linux that requires no account and no subscription.

The project listing dated Oct. 8, 2026 describes a full stills workflow built around local files and local compute, according to Rembrandt on GitHub. No login is required.

Editing is non-destructive. Adjustments are saved as XMP files stored alongside the photos, so the originals are left unchanged. The sidecar files act like instruction notes in ordinary folders, portable and compatible with file-based backup.

RAW support is stated as broad. Rembrandt supports RAW files, the unprocessed data from the camera, from more than 1,000 cameras with GPU-based development. The development path runs on the graphics processor rather than the CPU.

A word-based editing mode runs entirely on-device. It uses a built-in vocabulary, not a language model, with input mapped from a constrained set of terms inside the application.

Super Resolution is offered at 2x and 4x. It runs on the GPU, including Metal on Apple silicon, with the same local-acceleration approach for high-resolution output as for interactive editing.

AI Denoise is included for high-ISO shots, photos taken in low light at high sensitivity. It cleans the capture into a new DNG. The application can also merge HDR brackets, panoramas and focus stacks into RAW-like DNG files. Both functions produce DNG intermediates rather than rendered pixels alone.

Masking covers manual and AI-assisted selection. The set includes brush, gradients, colour and tone ranges, and AI subject, background, object and depth. The list allows selective correction without leaving the non-destructive workflow.

The broader context here is control of the pipeline. Local XMP sidecars, local GPU development and local vocabulary matching keep the edit loop on the machine. For privacy-sensitive work and for field work with unreliable connectivity, that locality removes entire classes of failure and disclosure risk. It also reduces latency variance, since sliders and mask updates do not wait on a round trip. Files remain browsable and copyable with standard tools, and edits travel as small text beside them.

In my view, the DNG-centered design is the detail to weigh most carefully. Denoise output and merged HDR, panorama and focus-stack output are described as DNG or RAW-like DNG, which preserves a flexible stage for later tone and color decisions. Keeping Super Resolution, denoise and merges inside one GPU-accelerated application shortens the chain of helpers and converters that often accumulates around difficult files. Worth flagging, constrained-vocabulary editing is a different trade from open-ended prompt editing. It should be judged on precision and repeatability for known corrections, not on conversational breadth.

Looking at what this means for daily work, cross-platform availability with Metal acceleration on Apple silicon lowers friction for mixed-OS teams and for photographers who move between desktop and laptop. Support for more than 1,000 cameras in one development path reduces the wait for new-body support that can stall ingest. If the implementation holds up under real libraries, the practical gain is straightforward. More keepers from high-ISO shoots, more usable brackets and stacks, and larger printable files, all without adding a subscription or an account to the workflow.