Commentary#Photorealistic#Image editing
LightCraft tested: the free Lightroom alternative, and the color problem inside it
LightCraft is a free, open-source Lightroom alternative in Rust, self-rated 49% for a working photographer. Color matrices decide if it fits your workflow.
LightCraft is an open-source Lightroom alternative rebuilt from scratch in Rust by the ArtCraft team: a photo library and raw developer whose entire pipeline is scene-referred, wide-gamut and 32-bit float, with originals never touched. As of October 12, 2026 it has 10,686 GitHub stars (2,985 when we published the seven-app comparison on October 8), and the latest release, v0.6.0, shipped on October 11. Its self-ratings are the interesting part. By feature count it claims about 81% of Lightroom — 98% of core features — but as a day-to-day Lightroom replacement it puts a typical working photographer at roughly 49%. That 32-point gap is the subject of this article: LightCraft’s darkroom is genuinely strong, camera color calibration is its acknowledged weak spot, and in photography the workflow succeeds or fails on color accuracy.
Key points
- Two self-ratings, 32 points apart: about 81% of Lightroom by feature count and 98% of core features, against about 61% for the product overall, 49% for a typical working photographer and 64% for casual use. A complete feature checklist does not equal a workflow that runs.
- Color matrices are the project’s admitted biggest gap: Sony, Nikon, Panasonic, Fujifilm and Canon CR3 raw color comes from “guarded estimates” derived from camera JPEGs, with built-in profiles for the ILCE-7CR, ILCE-7M4, X-H2S and X-T4; measured calibration is missing and everything else keeps a neutral matrix. That is why the same frame can look different from Lightroom.
- The darkroom and the speed are real: 18 presets, an 8-band color mixer, three-way grading wheels, 15 local adjustments per mask; full-size export in about 0.3 seconds, a 24 MP raw rendered in 0.2–0.5 seconds, and stepping through a shoot at roughly 50 ms per photo. On speed it does not behave like an alpha project.
81% and 49%
The feature status section (as of 2026-10-10) gives a feature count first and a workflow verdict second. Against Lightroom’s feature list it is about 81%, with core features at 98%, tracked row by row in docs/parity-checklist.md. As a day-to-day Lightroom replacement, the whole product sits nearer 61%, a typical working photographer nearer 49%, and casual use nearer 64%. The project’s own scope statement is specific: on a single machine it handles JPEG, DNG, and most Sony, Panasonic, Fujifilm, Nikon and older Canon raws, with the per-camera detail in docs/raw-parity.md.
Ten days ago, at the seven-app comparison, the figures read “feature count about 79%; day-to-day replacement about 60–70%.” The framing changed: the old single range for day-to-day use became three tiers, and the 49% now names the group that matters commercially. Splitting a number into segments usually means the measurement got more precise, and precision here moved the conclusion from “usable for most people” to “not usable for most working photographers.”
Why does 98% of core features not equal usable? Photography tolerates error very differently from image editing. Retouching one commercial composite is iterative — if the result is wrong you make another version. A wedding or product shoot produces hundreds of originals, and color that is off by a little means reworking the batch or correcting every frame by eye. LightCraft’s checklist can cover every line Lightroom has, but “covered” and “trustworthy output” are separated by an entire calibration chain.
Color is the soft spot
The README states this directly, which is rarer than it should be: camera color calibration is the biggest gap. Raw color for Sony, Nikon, Panasonic, Fujifilm and Canon CR3 comes from guarded estimates of the camera’s own JPEG, with built-in profiles for the ILCE-7CR, ILCE-7M4, X-H2S and X-T4, and no measured calibration yet; other raw formats, or fits rejected by the guard, keep a neutral matrix. That is the precise version of the claim that circulated a fortnight ago — “other raw formats fall back to a neutral matrix, so color looks muted.”
The impact varies enormously by photographer. Anyone shooting JPEG or DNG for screen delivery will barely notice: the pipeline is scene-referred and wide-gamut in 32-bit float, color is computed in OkLCh, a modern perceptual space, highlights roll off like film and shadows open without halos, so color stays clean from capture to export. Anyone whose work depends on the absolute color fidelity of a body raw will notice immediately — studio shoots, ID photos, e-commerce product imagery, anywhere the output has to match a physical object, a color off is a reshoot.
Community reaction matches that risk profile. A photographer on Hacker News loaded the same raw into Lightroom and LightCraft side by side and reported that the results looked completely different. Comparisons like that cannot isolate operating differences — different development parameters, different default settings — so as a first impression it is unreliable. The useful question is whether your image matches your camera’s own JPEG and the other mature open-source tools, not whether it matches Lightroom, whose output varies by version and defaults.
Three adjacent gaps belong in the same paragraph. Compressed Olympus raws and unsupported CR3 variants read only the embedded JPEG preview today, though Fujifilm’s lossless and lossy compressed RAF now decodes sensor data. AI masks and denoise are classical heuristics rather than learned models. Network libraries read only the raw header and embedded JPEG — 56.5 MB down to 5.9 MB and 10 seconds down to 1 second on the test share — but the developed picture still needs the sensor data over the wire. On-screen HDR display, the video modules, and Lightroom’s Classic Print, Book and Map modules are all absent.
If color accuracy is the thing you cannot compromise on, there is a practical way to get confidence before committing. Shoot a frame of a color checker chart or a known object under your studio lights, develop it in LightCraft with neutral settings, and compare it against both the camera’s own JPEG and the same frame in darktable. Three comparisons tell you what the self-rating cannot: whether your body is on the built-in profile list, whether the neutral matrix lands close enough for your delivery standard, and whether any of the difference is your expectation rather than the software. Photographers with a repeatable test frame keep that frame for exactly this reason.
The darkroom itself is solid
Setting calibration aside, LightCraft’s development tools belong at the top of the open-source field. The Light panel is exposure, contrast, highlights, shadows, whites and blacks, where highlights use edge-aware local tone mapping on log-luminance: pulling highlights to −100 recovers a blown sky without the grey halos a naive curve leaves behind. The Color panel has white balance (Kelvin for raw, relative for JPEG, with presets, Auto and a click-to-neutralize eyedropper), skin-protecting vibrance, saturation, an 8-band color mixer across hue, saturation and luminance, and three-way color grading wheels with blending and balance. The Effects panel is texture, clarity, dehaze (dark-channel prior with guided refinement, negative values adding atmosphere), a post-crop vignette with highlight priority, roundness and feather, and resolution-independent film grain. The tone curve is a parametric region curve plus point curves for RGB, red, green and blue, monotone by construction so you never hit an accidental tone inversion.
Masking is one of the most complete sections in the list: a brush with size, feather, flow, density and erase; linear and radial gradients with draggable pins; selection by luminance range, color range, sky, subject and background; components combined with add, subtract and intersect, any of them invertible. Each mask carries 15 local adjustments — temperature, tint, exposure, contrast, highlights, shadows, whites, blacks, texture, clarity, dehaze, hue, saturation, sharpness and noise — plus an overall amount. The README’s example stacks a linear Sky mask under a radial Sun glow, which in Lightroom usually means a plugin or a long trial-and-error session.
The preset box ships 18 hand-built presets (Golden Hour, Teal & Orange, Faded Matte, Selenium Tone, Crisp Landscape and more), each with an amount slider from 0 to 200 percent; your own settings save as new presets, can be favorited, and can be copied, pasted or synced across a whole selection in exactly the groups you choose. One click (V) goes monochrome with an 8-band B&W mix to darken skies or make foliage glow, after which the grading wheels add selenium, sepia or a split-tone print. Before-and-after is covered too: hold \ to peek at the original, Y for side by side, Shift+Y for a split. Crop and geometry offer free or locked ratios (1:1, 4:5, 5:7, 2:3, 4:3, 16:9, 16:10, original), drag-to-rotate straightening that keeps the largest crop inside the frame, thirds, grid and golden-ratio overlays, flips, 90° rotations, and manual vertical, horizontal, rotate, aspect, scale and offset transforms.
Fast, native, private
The performance numbers do not read like an alpha project. A full-size export takes about 0.3 seconds including a parallel JPEG encode. Opening a raw shows its embedded camera preview or cached render within about 0.1 seconds, with the full render following at 0.2–0.5 seconds for a 24 MP file. The next and previous photos are prepared in the background, so stepping through a shoot runs at roughly 50 ms per photo. The loupe, before-and-after views and every visible thumbnail render on a worker pool off the UI thread — drafts during drags, full quality on release. With no GPU, the same pipeline runs across all CPU cores and redoes only the stages a given slider affects; batch exports render several photos side by side so one photo’s decode and encode overlap another’s render.
The privacy story is equally clean: no account, no cloud, no telemetry, no subscription. The catalog is an append-only log of human-readable operations that you can diff, back up or replay. For sensitive work — client portraits, ID documents, unreleased products — “the originals never leave this machine” is a selection criterion on its own.
The engine underneath explains the speed numbers. LightCraft is pure Rust with no C in the pipeline, and it ships its own raw decoders — DNG, Canon CR2/CR3, Sony ARW, Nikon NEF, Fujifilm RAF including X-Trans, Panasonic RW2 and Leica RWL, Pentax PEF, Olympus ORF — plus its own color science and its own pipeline. Internally that means linear Rec.2020 float with Bradford-adapted white balance and gamut mapping. JPEG, PNG, TIFF, WebP, PSD composites and JPEG XL open today, and HEIC/HEIF photos open with the heif build feature, where HEVC itself is a build-time choice just as it is in PhotoCraft. Owning the decoders is a maintenance burden the project accepted; it is also why a raw opens in tenths of a second instead of handing off to an external library.
Library and search
The library stays out of the way: All Photos, Recently Added, Picks, By Date, Albums nested inside Folders, and Recently Deleted. Rate with 0–5, flag with P / X / U, color-label with 6–9. Search understands fields — rating:>3 flag:pick iso:>800 camera:x2 date:2026-04 keyword:mountains — and composable field queries are unusual in an open-source library, which pays off when culling a large shoot. Every view sorts by capture date, import date, edit date, name, rating, size or a stable random shuffle, and the justified Photo Grid and Square Grid views are virtualized.
This layer is the underrated half of Lightroom: half its value is culling efficiency, half is applying settings in bulk. LightCraft puts both inside one native application, and because the catalog is an append-only log, a bulk preset application stays traceable and reversible.
It also ships a procedurally generated demo library, which is more useful than it sounds: a new install has real photographs to cull and develop before you point it at a client’s shoot, so the culling workflow and the field search syntax are learnable on material that carries no risk. The README’s masking screenshots use that library, and the before-and-after comparisons use public-domain frames — Ansel Adams from the U.S. National Archives, Dorothea Lange from the Library of Congress, NASA Apollo 8 and Apollo 17 imagery — developed inside the app. Every screenshot in the documentation was captured automatically by an agent through the control channel rather than by hand, which is a small but telling signal about how this project is built.
LightCraft vs darktable vs RawTherapee
The free raw-development seat in 2026, with facts checked on 2026-10-12:
| Dimension | LightCraft v0.6.0 | darktable 5.6.2 | RawTherapee 5.13 | Lightroom (single app) |
|---|---|---|---|---|
| Price | free, MIT / Apache-2.0 | free, GPL-3.0 | free, GPL-3.0 | $11.99/month (US, 2026-10-07) |
| Interface | mirrors Lightroom | its own, technical | its own, technical | — |
| Own raw decoders | yes (DNG, CR2/CR3, ARW, NEF, RAF, RW2/RWL, PEF, ORF) | external libraries | external libraries | yes |
| Color science | own, Bradford-adapted white balance, linear Rec.2020 float | mature | mature | commercial-grade |
| Camera calibration | four built-in profiles, rest estimated | mature measured community calibration | moderate | most complete |
| Learning curve | low (the layout is Lightroom’s) | steep (one review estimates two to four weeks of deliberate practice) | steep | low |
| GPU | yes | yes | limited | yes |
| Privacy | no account, no cloud, no telemetry | same | same | cloud-connected |
| Self-rated | alpha, ~49% for a working photographer | stable | stable | commercial software |
The table sums up in three sentences. darktable’s raw handling is mature and the price is time — Chuck Henrich’s estimate is two to four weeks of deliberate practice. RawTherapee has its own strengths in demosaicing and detail recovery. LightCraft has the lowest migration cost and the cleanest privacy story, while its calibration column stays unfilled today. Lightroom’s single-app price of $11.99 a month is half of the $22.99 that Photoshop costs, which is part of why its subscription draws less resistance.
The demand signal is strong on YouTube: photography educator Anthony Morganti’s “The New FREE Lightroom Clone Everyone Is Talking About: LightCraft!” has 30,701 views and Tom Calton’s “8 Lightroom Alternatives That Will Save You Money!” has 414,921 (both read 2026-10-12). Lots of people want out of Lightroom; whether they can depends on their raw format and their color requirements.
Choosing between the two is less about features than about where you spend your attention. darktable asks you to learn a different way of thinking about the image — the develop module, the film emulation, the sidecar workflow — and repays that investment with a mature calibration community and an enormous effect library. LightCraft asks you to learn almost nothing, because it assumes you already know Lightroom, and repays it with speed and a catalog you can script. If your learning budget is zero, LightCraft wins; if your color budget is high, darktable wins. The common mistake is picking on the basis of feature checklists, which both projects publish and neither of which measures accuracy.
The door left open for agents
Every menu item, slider, brush stroke, crop handle and keystroke is a command with a stable id and JSON parameters, and the UI, the keyboard, the CLI, a JSON-lines control channel and an MCP server all dispatch through the same entry point. An agent can cull a shoot, develop it, mask a sky and export it — and see the result; every demo screenshot in the README was captured automatically by an agent through that control channel.
For photography this interface is more useful than it is for image editing, because “cull and first-pass develop an entire shoot” becomes a repeatable pipeline. Set rating and ISO thresholds, let the agent pull candidates with field search, apply presets, and export grouped by camera body; a human does the final pass. For library-scale delivery — real estate, e-commerce, events — that is worth trying today. The bitmap-side equivalent is covered in our PhotoCraft test.
Who should try it, who should wait
- Personal photographers shooting JPEG or DNG for screen delivery: try it now. The interface is Lightroom’s, the darkroom is complete, there is no account and no cloud, and it is fast — migration costs almost nothing.
- Workflows that need absolute color consistency from body raw (studio, e-commerce, ID photos): wait for measured calibration. Outside the four built-in profiles your body will most likely land on the neutral matrix, and color drift becomes a batch rework.
- Experienced darktable or RawTherapee users: no strong switching case — those projects lead on maturity and calibration communities. LightCraft works well as a second opinion or a quick-export tool.
- Automation and bulk work: worth trying. The MCP surface plus the CLI can script culling, presets and export.
Whichever group you fall into, the test is cheap: import the two dozen frames from a shoot you have already delivered, develop them with the sliders you would have used anyway, and compare the results against your Lightroom export at full size. The color question that the self-ratings cannot answer takes twenty minutes to answer on your own files — and if the answer is yes, the remaining gaps are modules you may never open.
LightCraft belongs to a bigger picture: our seven-app comparison weighs all of ArtCraft’s Adobe alternatives. The GitHub repository updates daily; downloads live at getartcraft.com/apps/lightcraft.
FAQ
Is LightCraft free for commercial use?
Yes. MIT or Apache-2.0 dual license, with no account, no cloud, no telemetry and no subscription; the catalog is a local append-only log. Licensing of your photographs is your business, and the software license stays out of it.
Does it know my camera?
Most of them. The decoders cover DNG, Canon CR2/CR3, Sony ARW, Nikon NEF, Fujifilm RAF including X-Trans, Panasonic RW2, Leica RWL, Pentax PEF and Olympus ORF. The project’s stated scope is JPEG, DNG, and most Sony, Panasonic, Fujifilm, Nikon and older Canon raws on a single machine, with a per-camera list in docs/raw-parity.md. Compressed Olympus raws and unsupported CR3 variants read the embedded JPEG preview for now.
Why doesn’t the color match Lightroom?
Different calibration data, not a different pipeline. Outside its four built-in profiles, LightCraft’s raw color comes from guarded estimates of the camera JPEG and falls back to a neutral matrix, while its rendering pipeline — scene-referred, 32-bit float, OkLCh — is solid. So the test worth running is whether your image matches your camera’s own JPEG and darktable, not whether it matches Lightroom, whose output varies with version and defaults.