Camera Raw

A camera raw file — .dng, .cr2, .cr3, .nef, .arw, and others — is not a picture yet. It is the sensor's own measurements, saved before the camera (or any software) decided what they should look like. Paintapp opens raw files by running that decision itself: demosaicing the sensor mosaic into an image, then applying the same white-balance, camera-calibration and tone pipeline that Adobe Camera Raw and Lightroom use, so the result looks like a photograph rather than a strange colored grid — and so the same sliders and camera profiles behave the way they do in those tools.

This is a work in progress. Raw decoding, color, and the develop sliders are all actively being built out. This page describes what exists today as accurately as possible, including the rough edges — see Current Limitations.

What Makes Raw Different

A JPEG or TIFF straight off a camera has already been through a full render: demosaiced, white-balanced, color-corrected, tone-mapped, sharpened, and compressed — a process the camera's own processor completed in a fraction of a second, discarding everything it didn't keep. A raw file skips all of that. Each pixel position holds one measurement — red, green, or blue, never all three — arranged in a checkerboard-like mosaic called a Bayer pattern (or, on some Fuji sensors, an X-Trans pattern), plus enough metadata to reconstruct what the camera saw: the black level, the sensor's saturation point, and the color of the light in the scene.

Nothing is thrown away and nothing is decided for you. That's the appeal of raw — you (or your software) choose the white balance, the tone curve, and the color rendering, instead of living with the camera's one-shot guess — but it also means a raw file needs a real pipeline before it looks like anything.

Opening Raw Files

Recognizing a format and fully decoding its sensor data are two different milestones, and they don't land together — a vendor's compression scheme has to be reverse-engineered and implemented one at a time. Today:

Format	Vendor	Status	Notes
DNG	Adobe (universal)	Full color	Both uncompressed and lossless-JPEG-compressed DNGs decode end to end. A DNG carries its own sensor calibration in the file, so this is the most accurate and most reliable path in.
CR2	Canon	Approximate (5D3 accurate)	Sensor data decodes correctly (geometry exact), and the shot's own white balance is read from Canon's maker notes, so the picture opens correctly balanced. A CR2 carries no color calibration, so that part is approximate except on the EOS 5D Mark III, which has a bundled calibration table — see Camera Profiles below for how this gets fixed per camera.
CR3	Canon	Real white balance	Canon's current raw format, and a different kind of file from the ones above: not a TIFF at all, but the same container a video uses, with the sensor data stored as its own `crx` codec. Both variants open — the full-size RAW and the smaller lossy C-RAW. The picture is cropped to the sensor's active area and balanced with the shot's own white balance, read from where Canon actually keeps it; only the camera's color calibration matrix is still estimated.
NEF	Nikon	Approximate (D300 accurate)	Nikon's Huffman-compressed sensor data decodes via an in-house decoder, and the shot's own white balance is read from Nikon's maker notes. Color calibration is approximate except for the Nikon D300, which has a bundled table — see below.
ARW	Sony	Approximate color	Both of Sony's compressed sensor formats decode: the older fixed-rate CFA pack used by most bodies, and the newer tiled-lossless-JPEG scheme used by recent A-series cameras. Sony's own metadata isn't read yet, so both the white balance and the color calibration are estimated rather than taken from the file.
RW2 / RWL	Panasonic / Leica	Real white balance	Panasonic's packed sensor format decodes via an in-house decoder. Unlike the vendor formats above, an RW2 keeps its black/white levels and the shot's own white balance in ordinary file tags, so the picture opens correctly balanced rather than guessed; only the camera's color calibration matrix is still missing.
PEF, SR2, NRW, SRW, ERF, DCR, KDC, 3FR	Pentax, Sony, Nikon, Samsung, Epson, Kodak, Hasselblad	Not yet	Recognized, and the uncompressed and lossless-JPEG files among them decode. Where the camera used its own compression, opening one reports a clear "unsupported compression" error rather than showing a broken or blank image.
RAF, ORF	Fuji, Olympus	Not yet	Not opened at all yet: these are always written with the vendor's own sensor compression, so rather than advertise a format that could only ever fail, they are left unregistered until their decoders land.

Demosaicing

Turning the Bayer mosaic into a full-color image — filling in the two missing channels at every pixel by looking at its neighbors — is called demosaicing. It is worth knowing what this step really is: two thirds of the color was never measured, so it is reconstructed from the surrounding pixels. Every raw converter is making an informed guess here, and there is no single correct answer to compare against.

Paintapp uses AHD (Adaptive Homogeneity-Directed interpolation), which chooses a direction to interpolate along for each pixel rather than blindly averaging, so it holds fine detail and avoids the colored fringes and maze-like artifacts that show up on tight patterns like fabric, brickwork or distant foliage.

Because the detail is reconstructed rather than recorded, the same raw file can resolve slightly differently in different software — your camera's own converter, Lightroom and Paintapp may each render a hair, a roof tile or a fence line a little differently at 100% zoom. That is normal and not a fault in any of them; they are simply different readings of the same missing data. It only shows at the finest scale — overall color, tone and sharpness come from the stages below, which are shared.

From Sensor to Picture

Once the mosaic is demosaiced, Paintapp runs the same chain Adobe Camera Raw runs, as a sequence of ordinary nodes in the document's node graph — visible and editable like any other filter, not a hidden black box:

Stage	What it does
Demosaic	Reconstructs full-color pixels from the sensor mosaic, working in the camera's own linear light — nothing is white-balanced or color-corrected yet.
White Balance	The Temperature and Tint sliders. White balance isn't a simple color tint applied afterward — it changes which part of the camera's dual-illuminant calibration gets used (see Camera Profiles), so it has to run first, before color is fully resolved. "As Shot" reproduces what the camera itself decided at capture time; "Custom" lets you pick a Kelvin temperature and tint by hand.
Camera Profile	Applies a `.dcp` camera profile's "look" — its hue/saturation adjustments and base tone curve — on top of the calibrated color from the previous stage. This is where "Adobe Standard," "Camera Vivid," "Camera Landscape," and similar profile names come from.
Basic Tone	Exposure, Contrast, Highlights, Shadows, Whites, and Blacks — the familiar Camera Raw / Lightroom Basic panel. On a raw document this stage also renders the default picture before any slider moves, using the same rendering steps as Adobe's own DNG SDK, so a raw with every slider at zero already looks like a normal photo rather than something flat and dim.

If a raw file was already edited in Lightroom or Adobe Camera Raw, those edits travel with the file as XMP metadata (the crs: fields — white balance, exposure, contrast, highlights, shadows, whites, blacks, and the chosen camera profile name). Paintapp reads that metadata on open and seeds the sliders from it, so a raw you already developed elsewhere opens showing those same edits instead of resetting to defaults.

None of this pipeline runs on ordinary images — every one of these nodes checks whether its input is actually raw sensor data (tagged with an XYZ profile by the reader) and passes non-raw pixels through untouched.

Camera Profiles (.dcp)

A .dcp ("DNG Camera Profile") is Adobe's format for describing exactly how one camera model's sensor should be rendered into color. It is not a fixed lookup table — it bundles several pieces that get combined at develop time:

Two color-calibration matrices (and their inverses), one measured under warm tungsten light and one under daylight — because a sensor's raw response to "this is a neutral gray" genuinely differs depending on the color of the light hitting it. Paintapp interpolates between the two based on the shot's actual white balance.
Hue/saturation shift tables (one per calibration illuminant) that nudge specific colors — the classic example is pulling reds toward orange, or deepening sky blues — the same kind of per-hue tuning a camera maker bakes into its own JPEG engine.
A look table and a base tone curve — the part that gives a named profile like "Camera Vivid" or "Camera Landscape" its character.

Adobe's own .dcp files are Adobe's intellectual property, so Paintapp does not ship or redistribute Adobe's per-camera profile library. What ships is the parser and the Camera Profile node. You supply the profiles — most people already have a large set on their machine, because Adobe's free DNG Converter, Lightroom and Camera Raw each install one.

There are two ways to use them:

Load one by hand — pick a .dcp file on the Camera Profile node. Good for trying a specific look on a specific shot.
Let it match automatically — copy your .dcp files into the app/color/dcp_profiles folder (you can drag them there in the File Browser). From then on, opening a raw picks the profile for that camera on its own: the one the file was last edited with if it names one, otherwise that camera's Adobe Standard.

With no matching profile the node simply passes through, and the picture still renders from whatever calibration the raw file — or the built-in table below — provides.

Where the table above says color is "approximate", that is what's missing: unlike a DNG, a CR2, CR3, NEF or ARW doesn't carry its camera's color calibration inside the file, because the vendor keeps it separate — which is what a .dcp profile supplies. White balance is a different thing and is not missing: for CR2, CR3, NEF and RW2 the shot's own white balance is read straight from the file, so those already open correctly balanced.

A couple of bodies — the Nikon D300 and Canon EOS 5D Mark III — have their calibration built in, so they render with accurate color and a fully working White Balance slider without any profile at all. For every other camera, loading a matching .dcp is what turns "approximate" into "accurate".

Camera Profile vs. ICC Profile

It's a fair question: Paintapp already has a full ICC profile system — why does raw need a whole separate format instead of just an ICC profile? Because they answer different questions.

	Camera Profile (.dcp)	ICC Profile
What it describes	How to turn raw sensor readings from one specific camera model into color	How one specific device (a monitor, printer, or already-rendered image) relates to a device-independent reference space
Depends on lighting?	Yes — it carries two calibrations, for two different light sources, interpolated by the shot's white balance	No — one fixed transform, regardless of what the image depicts
What it operates on	Sensor data that has no color yet (a raw pixel is one brightness measurement, not an RGB triplet)	RGB (or CMYK/Lab) values that are already real colors
When it applies	Once, right after demosaicing, as part of turning sensor data into a picture	Any time afterward — for display, print simulation, conversion between color spaces, or handing the file to another application

The short version: a raw pixel isn't a color yet, so there's nothing for an ICC profile to describe. A .dcp camera profile does the job a camera's own processor does when it turns sensor data into a JPEG — decide what the sensor's numbers actually mean, accounting for the light they were captured under. Once that's done — once Camera Profile and Basic Tone have produced an ordinary picture — the result is just an image like any other, and Paintapp's regular ICC (or OCIO) pipeline takes over from there for display, proofing, and export, exactly as it would for a JPEG or PSD. They're stages in a sequence, not competing options.

Where ICC and OCIO Fit In

Put together, opening a raw file flows like this:

Step	Color space
Sensor mosaic → Demosaic	Camera-native linear light
White Balance → Camera Profile → Basic Tone	Scene-linear CIE XYZ, internally by way of linear ProPhoto — working color, not yet meant for a screen
Display / export	Handed to Paintapp's existing ICC / OCIO pipeline, the same one every other document uses

The document stays tagged with a device-independent XYZ profile throughout the raw-develop chain. Nothing about the Camera Raw nodes is aware of your monitor, your proof setup, or an OCIO config — that separation is deliberate, and it's why the same color-management settings and soft-proofing tools documented on the Color Management page apply to a developed raw file exactly as they do to any other document.

Browsing Raw Files

The File Browser thumbnails raw files the same way it thumbnails any other image — no separate raw-specific browsing mode. If you keep your shots in a folder on disk, granting Paintapp access once (as a Local Root Folder) lets it reopen that access automatically the next time you're back at the same location, which is covered in the File Browser page along with the rest of its folder handling.

Current Limitations

A few vendor formats aren't decoded yet — Fuji's X-Trans RAF and Olympus ORF aren't opened at all, and the vendor-compressed variants of Pentax PEF and a handful of other TIFF-based raws report an unsupported-compression error. See the table above.
Vendor color calibration varies. Outside of DNG (and the two built-in calibrations, for the Nikon D300 and Canon EOS 5D Mark III), CR2/NEF/ARW/RW2 files use an approximate, camera-generic calibration rather than that camera's real one, unless you supply a matching .dcp yourself. White balance is separate and is read from the file for CR2, CR3, NEF and RW2 — only ARW still falls back to a guess there.
No camera profile library is bundled. Adobe's per-camera .dcp files are Adobe's IP and aren't shipped with Paintapp — accurate, camera-specific "looks" require loading your own .dcp (see Camera Profiles above).
The White Balance Temperature slider has a known bug — it currently has little to no visible effect at mid and high Kelvin values, and behaves incorrectly at low values. This is under active investigation.
The Basic Tone sliders are a mix of validated and provisional curves. Highlights and Shadows have been recaptured to match Photoshop's actual behavior closely; Exposure, Contrast, Whites and Blacks are still on an earlier, less precisely matched curve set pending the same treatment.
No writing raw develop settings back out. Paintapp reads a file's existing Lightroom/Camera Raw XMP edits on open, but doesn't yet save your own edits back to an XMP sidecar or export a develop-settings preset.