Darktable is a photography catalog and raw-image developer. It imports references to camera files into a library, lets the user sort and rate them in the lighttable, then sends a selected image through a chain of processing modules in the darkroom. Darktable does not paint those adjustments into the camera original. It records the instructions in a library database and an XMP sidecar file, then creates a finished JPEG, TIFF or another output during export. This is different from opening a JPEG in a basic editor and saving over the same pixels.
The library holds context
An image must enter Darktable’s library before the program can edit it. Importing usually records the source path and creates an XMP sidecar beside the original; it does not automatically gather every photo into a managed folder. Moving or renaming a source folder outside Darktable can therefore leave its library entry unable to locate the file.
Ratings, color labels, tags, collections and editing history help narrow a large shoot without changing the raw capture. The database keeps this information for quick access. Sidecars carry most of the same image-level work so another Darktable library can reconstruct it. A backup that contains only the raw files omits those decisions, while a backup that contains only the database omits the photographs.
Sidecars need ownership
Darktable treats its database as the authority after an image has been imported. If another program changes the XMP sidecar behind Darktable’s back, the running library does not automatically adopt that edit. Darktable may overwrite the external change the next time it synchronizes its own state. A startup preference can check for changed sidecars and ask which copy should win.
This conflict matters when two computers share one photo folder. Each library can carry a different edit history for the same raw file. Synchronizing the folder does not merge those histories. The user must decide which sidecar becomes authoritative before both systems continue writing.
Its duplicates are virtual versions. They share the same input image but have separate sidecars and processing histories. Deleting the raw source removes the common foundation for every duplicate. Custom sort sequence values also do not travel in sidecars, so a rebuilt library may not reproduce a hand-arranged sequence.
Modules form a pixelpipe
Darktable calls its ordered processing chain the pixelpipe. Each active module receives the previous result, performs its operation, blends through any mask, and passes output onward. The displayed module order is the execution order. Moving a module can change the final image even when every slider keeps the same value.
Drawn and parametric masks limit a module to selected areas or pixel properties. A mask can depend on the module input or a processed result, so changing an earlier module can also change which pixels the later mask selects. This relationship makes a complex edit reproducible, but it means a small upstream change may have consequences farther up the stack.
Preview pipelines use reduced work for interactive speed. Export uses the full-size image at output quality. Fine noise, sharpening and mask edges can therefore look different at screen zoom levels that do not map one image pixel to one display pixel.
Export makes the photograph
Darktable must export an image before another viewer can see the developed result. The export module decides destination, filename template, dimensions, color profile and format. Its common file-on-disk route can create a darktable_exported directory beside the originals when the user leaves the default destination. A large batch may therefore consume space on the source volume rather than an intended delivery drive.
JPEG suits normal distribution but applies lossy compression. TIFF retains more data for continued work in another raster editor. Exporting either format does not replace the raw file or its sidecar. The final output, the original and the editing record remain three separate objects, and a complete archive must keep the parts needed for the intended future work.
