PhotoKoma

One straight road from a full memory card to finished photographs

A darkroom for a whole shoot rather than one photo: it pairs RAWs with their JPEGs, throws out the duplicates and the misses, groups what belongs together, and carries the look you give one photo to the rest of its group - working each frame out on its own rather than copying slider positions across.

What it does

ON THE WAY IN

Pairs and previews

RAW and JPEG pair by name, and the first previews come from the RAW's own embedded thumbnail rather than from developing it.

FIRST PASS

Sets aside the repeats

Bursts and the frames that missed go to a review pile with the reason attached. Nothing is deleted, and one click rescues.

READS THEM

Keywords every frame

The same pass that scores a photo also tags what's in it, so you can search your shoot without typing keywords first.

STRUCTURE

Groups into scenes

Photos taken together are grouped into scenes, and each scene picks out the best frame to edit first.

THE POINT

Carries the look

Edit that one frame and the rest of the scene follows - each photo adjusted for its own exposure and white balance, not just given your slider positions.

YOUR HAND

Learns how you edit

Show it RAWs you edited elsewhere alongside the JPEGs you exported, and it works out which of its own sliders come closest to your look.

THE DARKROOM

Develops the frame

Tone, colour, detail, denoise, retouch, masks and layers - a full panel that folds down to the parts you reach for.

SELECTION

Cuts the album

Name a number and get a cross-section that covers the whole shoot instead of eight versions of its best moment.

OUT

Delivers

Format, size, watermark, metadata and file naming - saved as presets.

It can learn your handBeta

Point it at RAWs you edited somewhere else, plus the JPEGs you exported from them. It works out which of its own sliders come closest to what you did, and saves that as a style you can apply to photos you haven't touched yet.

Each folder you teach it is kept separately. Switch one off and its influence drops out without being deleted; add another and the style shifts. What it learns is filed by subject, so the way you treat water won't be applied to a portrait.

Working out an edit from a finished JPEG is guesswork, and the app says so before you start. Often it lands close enough to use as it is; when it doesn't, you have a solid starting point rather than a blank slate.

  • WHAT IT LEARNED Where you tend to put each slider, and how consistent you are. Steady means it's worth applying; scattered means it depends on the photo.
  • HOW CLOSE IT GETS How far its render still sits from your export. What it cannot match is usually brushwork.
  • WHAT IT HASN'T SEEN Subjects it doesn't have enough examples of, named, so you know which folder to show it next.

In the develop panel

The camera's own rendering

Start from your camera's own colour, contrast and lens correction instead of a flat RAW.

Tone

Exposure, contrast, highlights, shadows, whites, blacks and midtones, with a tone curve over the top.

Colour

White balance, saturation and vibrance, plus colour bands you place anywhere on the wheel, each with its own hue, saturation and brightness.

Detail

Sharpening at the pixel scale, texture inside one range of tones, clarity across the whole picture, and dehaze for the distance.

Denoise

Three methods, from a quick pass to one that measures this photo's own grain in the RAW file before removing any of it.

Elements of the photo

Found for you, with a dial each rather than a mask to draw. Elements: water, earth and sky. People: skin, faces, eyes and blemishes. Animals: fur and feather, with the texture brought up rather than smoothed.

Retouch

Clone, heal and remove. Remove closes the hole with patches of the photograph itself. Each spot undoes on its own.

Masks and layers

Sky with no model needed, objects found by name, and masks you paint. Each layer carries its own sliders and its own switches.

The frame

Crop and straighten, and corner shading that darkens or lifts.

Why it exists

A shoot has a handful of scenes and a thousand frames. Only one of those numbers should be your problem.

The work after a shoot is repetition: the same rejection, the same crop, the same lift of the same shadows, frame after frame. Software that only knows how to edit one photograph makes you the one supplying the repetition.

So the unit of work here is the shoot. Culling, grouping and the look you settle on all apply at that scale, and the frame in front of you is the one the app thinks is worth your attention next.

Nothing leaves the machine

The models that measure, group, keyword and mask your photos all run on your machine, using the GPU if you have one. They're third-party models with their own licences, listed in the app. Cloud keywording is available, off by default, and needs your own API key.

Made in Rust, Tauri and Svelte Running local, offline models For Linux, Windows and Android Version 0.6.0