dancancer / RawLab
git clone https://github.com/dancancer/RawLab.git
v0.8.2 · today
RawLab iOS app and Sony-to-Fuji RAW/LUT processing tools
—
stars, 24h
—
stars, 7 days
+123
stars, 30 days · #1,905
123
★ total
9
forks · 4 open issues
Star history starts 2026-10-20; the chart appears after the second daily snapshot.
LanguageSwift
LicenseGPL-3.0
Created2026-09-27 (23d ago)
Last pushtoday
Watchers1
OwnerUser
Links
README excerpt
# RawLab English | 简体中文 Cross-brand RAW development and film looks, built around a shared C++ processing core with native Windows, macOS, Android and iOS editors. ## Overview RawLab develops camera RAW files, applies compatible Fujifilm film-simulation LUTs and custom looks, and exports finished photographs without changing the original files. It is an independent project, not an official Fujifilm product; film looks do not promise an exact match to in-camera JPEGs. ## Features - RAW development with exposure, camera-space white balance, tone, color and sharpening controls. - Ordinary opaque RGB8 SDR images: JPEG/PNG on all four clients, TIFF on macOS/iOS/Windows and HEIF on macOS/iOS. Raster input shares the editing/export core without the RAW scene curve or camera controls. Unsupported HDR, high-bit-depth, transparent and multi-frame inputs are rejected rather than silently converted. See the input contract. - New RAW edits start with embedded-preview exposure matching. One-shot Auto exposure and an eight-channel HSL mixer are shared across all four clients; existing saved exposure baselines are preserved. - Built-in film looks with 0-200% strength, compatible CUBE/RLOOK imports and neutral-versus-edited comparison. - Fourteen master recipes with style artwork and matching controls across all clients. Applying a recipe replaces six adjustments and all 24 HSL values; 0-200% strength scales the current recipe controls, HSL and LUT together. Saved edits and frozen batches preserve manual changes and per-target RAW WB. - Drag-only mobile film/recipe lists and desktop overflow buttons, plus adaptive iPad portrait/landscape layouts. iOS adjustment ranges, steps and highlight direction match macOS without changing older saved rendering values. - Linear-light wavelet denoising with separate luma, chroma and coarse-chroma controls, plus matching vignette and deterministic grain on all four clients. Metal, GLES 3.1 and D3D11 accelerate the shared denoising and effects kernels. - Per-photo local edit memory and resumable batch export with frozen settings, cancellation and failed-item retry, without modifying RAW files. - Photo information overlays, zoom/pan and original-size or long-edge exports. Desktop file browsers can reveal photos and apply confirmed current settings. - JPEG and 16-bit PNG export on Windows, macOS and Android; JPEG export on iOS. Rendered DNG file export is available in single and batch workflows on all four clients. It bakes the current adjustments, not reversible sensor RAW controls. See the rendered DNG contract and limits. Readable capture metadata is retained, while output size and orientation are updated. - Metal, Direct3D 11 and OpenGL ES processing where supported, with automatic CPU fallback. HSL runs in the platform photo GPU pipeline; RAW decoding and file encoding remain CPU operations. - A shared C++17 core, command-line tools and a desktop LUT preparation utility. - Desktop Agent JSON editing, c