C-Shop is a native, GPU-accelerated, layer-based image editor. Layers, masks, blend modes, non-destructive adjustments, filters, type and vector shapes — the things your hands already know — in a single binary that is on screen in about 400 ms.
Built to behave the way an image editor is supposed to behave, so there is nothing to unlearn. These are the parts people use every day.
Raster layers, nestable groups, fill layers, and 27 blend modes. Layer masks and clipping masks, per-layer opacity and fill opacity, locks, and drag-to-reorder.
All four boolean modes, feather, expand, contract, grow and similar. Select a colour wherever it appears, or fit a selection's edge to the one in the photograph, which is what hair needs.
Brush, pencil, eraser and clone stamp share their size, hardness, opacity and flow with dodge, burn, sponge, blur, sharpen and smudge — so all of them behave the same way. Tablet pressure, and brushes made from a selection.
Take texture from elsewhere and tone from where it lands, so a mark on a cheek comes out without leaving a disc of slightly-wrong brightness. Paint one region back to an earlier point in the history.
Curves, Levels, Exposure, Hue/Saturation, Colour Balance, Black & White and the rest — each as a dialog with a live preview and as a non-destructive layer, with a curve editor and a live histogram.
Blurs, sharpen, noise, distortions, stylise and a custom convolution. The preview shows real full-resolution pixels rather than a proxy, so you judge detail before you commit to it.
Text layers from your installed fonts, edited live on the canvas with a caret. Rectangles, ellipses, polygons, stars and Bézier paths with a Pen tool and boolean operations — all of it re-editable after the fact.
A layer that keeps the picture it was made from, so scaling and rotating never wear it out. Place one picture in several spots and correcting it is one correction rather than four.
Drop shadow, glows, bevel and emboss, satin, colour, gradient and pattern overlays, stroke — each with its own blend mode and geometry, and a shared global light.
ICC profiles kept from the files that carry them, CMYK opened as four real inks, sixteen bits a channel through to export, and a canvas that shows the document on the screen it is actually on — with soft proofing.
Free Transform, warp and puppet warp, content-aware scale that carves out the dull parts instead of squashing everything, and a perspective crop that undoes the angle you shot at.
A layered project format of its own, PSD in and out with layers and masks intact, camera raw (DNG), SVG as editable shapes, animated GIF and APNG as a frame per layer, plus PNG, JPEG, TIFF, WebP and more.
An optional pack adds seven neural networks that run on your own machine — nothing is uploaded, and nothing here needs an account. Install it or don't; the editor is the same either way.
Ask for the dog and you get the dog. One model finds objects and names them, another turns a click or a box into a mask, and what you are left with is an ordinary selection — so everything the editor already does with one still applies.
Select what you want gone and it is filled in with what was probably behind it. Nothing outside the selection is touched — not one pixel — so there is no seam to clean up afterwards.
It reads how far away everything is, works out which way each surface faces, and lights it from somewhere else — a low warm lamp from the side turns an ordinary afternoon into late in the day. The light stays on the subject rather than glowing off its edges.
A phone picture at high ISO turns a night sky into magenta-and-green confetti. A state-of-the-art deep learning model — SCUNet, a Swin transformer inside a UNet — takes it out and leaves the window frames, the roof lines and the street lights exactly where they were. Shown at one pixel to one pixel, so what you see is what you get.
Grow a picture four times over with the detail a bigger sensor would have recorded, rather than the softness that resampling gives you. The whole document grows with it — canvas, layers, masks and all.
Every window that previews on the canvas can be pushed aside, so the picture is the preview.
Seventeen named styles — noir, faded film, bleach bypass, golden hour, blueprint, watercolour, coloured pencil and more — each one scaling itself to whatever size of picture you hand it.
Free, open source, and yours. It builds from source in one command and needs no system packages beyond a working graphics driver.
# Build and run
git clone https://github.com/stubbb/c-shop.git
cd c-shop
cargo run --release
# Open a file straight away
cargo run --release -- photo.jpg
# Add it to the applications menu
./packaging/install-desktop.shEverything it installs goes under your home directory, so it needs no root. The GitHub repository has the current build, the full feature list and the documentation — it is always ahead of this page.
C-Shop is a personal tool in active development, built for its author's own work and pushed as it changes. There are no stability promises — though the project format carries a version and keeps opening files you have already saved.