DxO PhotoLab 10 retouches sensor dust with AI

After opening a photo folder, the software reads the image information from the Raw files, downloads matching correction profiles from the DxO Cloud and, if desired, retouches sensor dust with AI. The new tool automatically detects spots left by a dirty camera sensor, taking some of the tedious manual work off photographers’ hands.

Automatic correction starts as soon as the files are imported. PhotoLab processes lens distortion, lens softness, vignetting and color fringing based on the combination of camera and lens. For photos taken at high ISO settings, the software uses DeepPrime XD3 noise reduction, which, according to the heise test, recovers a surprisingly large amount of detail. This requires a graphics card with at least 6 GB of VRAM (graphics memory), such as an Nvidia RTX or AMD Radeon RX 6000.

Color control is also expanded. PhotoLab 10 uses color wheels similar to those found in video editing: hue and saturation can be adjusted separately for shadows, midtones and highlights. The wider “DxO Wide Gamut” color space is intended to render green tones better than AdobeRGB in particular. Exposure, color, details and geometry continue to be stored non-destructively in DOP data.

The limits become apparent with dust retouching. Individual sensor spots can be selected, but according to the heise test, the AI does not work perfectly in every case. Its practical benefit nevertheless remains high, since many corrections can be handled without a manual clone-stamp tool. The feature was already included in DxO PureRaw 6 and is now available directly in the Raw developer.

And, more concretely? Photographers with dust-prone sensors mainly save time when working on image series; users who shoot at high ISO settings also need a suitable graphics card. Those who work with Affinity can export images there directly from PhotoLab and bring them back. PhotoLab 10 is therefore not an automatic replacement for every retouching task, but it brings several recurring steps together in one place—with an AI feature whose results still need to be checked.