RapidRAW

A beautiful, non-destructive, and GPU-accelerated RAW image editor built with performance in mind.

RapidRAW is a modern, high-performance alternative to Adobe Lightroom®. It delivers a simple, beautiful editing experience in a lightweight package for Windows, macOS, Linux, and Android.

I started developing this project as a personal challenge when I was 18. My goal was to create a high-performance tool for my own photography workflow while deepening my understanding of React, WGSL and Rust.

• 2026-10-24: Rewrote dehaze with edge-aware haze estimation, added dual-illuminant RAW color matrices and gamut compression, improved denoise in shadows and highlights, and added arrow key grid navigation in the library
• 2026-10-23: Redesigned data management as a storage page with cache limits, rewrote noise reduction and sharpening, and fixed AI hangs on macOS
• 2026-10-22: Added split and side-by-side before/after compare modes, a global job tracker, AI face and text masks, new sky and foreground models, and non-destructive BM3D denoise
• 2026-10-21: Added GPU acceleration for AI models via DirectML and CoreML, and a background AI job queue that regenerates AI masks, depth maps and denoise on paste
• 2026-10-20: Added Immich integration thanks to @cl1x, non-destructive negative conversion, auto adjust strength levels, and color-preserving highlight reconstruction
• 2026-10-19: Optimized memory usage, GPU shaders and the preview pipeline for faster thumbnails, editing and exports
• 2026-10-18: Added non-destructive AI denoise with disk caching and seamless tile blending, updated Lensfun database, and optimized Linux RAW preview pipelines
• 2026-10-17: Added movable on-canvas center controls for vignette and centré adjustments, touch multi-selection mode for Android, and cleaned default-only sidecar file handling
• 2026-10-16: Parallelized RAW post-processing and pixel format conversion for faster image rendering
• 2026-10-15: Added auto-straighten powered by horizon and line segment detection, and added DPI settings for print exports

• 2026-10-13: Added individual visibility toggles to every adjustment tool in the editor panel
• 2026-10-12: Added optional RAW baseline exposure and highlight preservation compensation
• 2026-10-11: Embed keywords, ratings, and color labels as standard XMP metadata during export, alongside improved RAW EXIF writing
• 2026-10-10: Added {original_filename_number} token support to library batch renaming
• 2026-10-07: Added 3D Relight effect with surface normal generation and on-canvas aiming shapes, alongside new Atmospheric Fog adjustment and staged preview caches
• 2026-10-06: Added AI-free mode, rewritten whites adjustment, and improved shadow and highlight recovery with mid-scale detail reinjection
• 2026-10-05: Implemented log-space guided filter for edge-aware adjustments and integrated cloud API
• 2026-10-04: Added Kelvin white balance mode with area sampling, fine adjust mode for curves, pick/reject quick filters, and improved export presets
• 2026-10-03: Embed sRGB ICC profiles in exports, calculate HSL hue in perceptual color space & improve lens EXIF detection
• 2026-10-02: Added Dutch language support and improved continuous RAW magenta highlight correction
• 2026-09-29: Optimize GPU texture uploads to reduce memory overhead and avoid unnecessary buffer copies
• 2026-09-28: Added Tool Focus mode to isolate and expand only active adjustment tools
• 2026-09-27: Tone curves now display inactive channels as colored overlay curves and reference markers
• 2026-09-25: Streamline adjustment panel customization and improve default tool visibility
• 2026-09-23: Added borders and aspect ratio padding to export options, plus reorderable and collapsible adjustment sections
• 2026-09-22: Added Apple RAW 9 decoding/denoising support and custom savable crop aspect ratios
• 2026-09-21: Implemented cancellable AI tasks with real-time UI tracking and backend preparation for cloud inpainting
• 2026-09-20: Center-anchored crop resizing when holding Ctrl/Cmd on crop handles
• 2026-09-19: Added center mark composition guide overlay to the crop tool
• 2026-09-17: Rewrite vibrance & local contrast preserving highlights adjustment
• 2026-09-16: Add highlights color reconstruction & improve exposure shader

• Showcase & Edits
• The Idea
• Key Features
• Supported Formats, Lenses & Languages
• Current Priorities
• AI Strategy
• Initial Development Log
• Getting Started
• Camera Tethering
• Command Line Interface (CLI)
• Companion Apps & Ecosystem
• System Requirements
• Contributing
• Special Thanks
• Support the Project

---

Showcase & Edits

Watch RapidRAW in action:

https://github.com/user-attachments/assets/4b9b932a-b18c-49d3-849f-e44d3bef0d12

Explore example edits processed entirely within RapidRAW. You can download the .rrdata sidecar file for each image to inspect or apply the exact adjustment parameters.

If you like the showcased images and want to see more of my own photography, check out my Instagram: @timonkaech.photography

The Idea

Key Features

Supported Formats, Lenses & Languages

RapidRAW supports a wide array of camera RAW formats, non-RAW image files, and automatic lens corrections.

RapidRAW relies on dnglab / rawler for RAW file decoding. You can check out the full list of specific camera models on the dnglab Supported Cameras List.

Brand / Manufacturer	Supported Extensions	Description
Adobe	.dng	Digital Negative
Apple	.pro	Apple ProRAW
ARRI	.ari	ARRI Raw
Canon	.crw, .cr2, .cr3	Canon Raw 1, 2, 3
Casio	.bay	Casio Raw
Contax	.raw	Contax Raw
Epson	.erf	Epson Raw
Fujifilm	.raf	Fuji Raw (X-Trans & Bayer)
Hasselblad	.3fr, .fff	Hasselblad Raw
Kodak	.kdc, .k25, .dcs, .dcr	Kodak Raw formats
Leaf / Imacon	.mos, .iiq	Leaf / Imacon Raw
Leica	.rwl	Leica Raw
Mamiya	.mef	Mamiya Raw
Minolta

Note: Is your camera's RAW file unsupported? Please open an issue directly with dnglab to request RAW decoding support for your camera model and upload a CC-0 licensed sample set to RawDB (see Contributing RAW samples). Once added there, support will be updated in RapidRAW.

RapidRAW isn't just for RAW files! You can also import, edit, and convert standard image formats:

• Standard Formats: .jpg, .jpeg, .png, .webp, .jxl, .tiff, .tif, .bmp, .gif
• High Dynamic Range (HDR) / Wide Gamut: .exr, .hdr
• Graphics & Textures: .tga, .ico, .dds
• Specialist Formats: .qoi, .ff
• Netpbm Bitmaps: .pnm, .pbm, .pgm, .ppm, .pam

RapidRAW supports automatic lens profile detection, distortion, transverse chromatic aberration (TCA), and vignetting correction powered by Lensfun.

Note: If your lens profile is missing, please open an issue directly on the Lensfun repository to request profile addition.

RapidRAW is fully translated into the following 15 languages:

• 🇬🇧 English
• 🇪🇸 Català
• 🇨🇿 Čeština
• 🇩🇪 Deutsch
• 🇪🇸 Español
• 🇫🇷 Français
• 🇮🇹 Italiano
• 🇳🇱 Nederlands
• 🇵🇱 Polski
• 🇵🇹 Português
• 🇷🇺 Русский
• 🇯🇵 日本語
• 🇰🇷 한국어
• 🇨🇳 简体中文
• 🇹🇼 繁體中文

Current Priorities

Here is an outlook on what is actively being developed and planned for the coming months:

• Astrophotography Stacking: Implementing GPU-accelerated star alignment and stacking (mean, median, sigma-clipping) to merge multi-exposure night sky sequences into 32-bit linear composites.
• Bug Fixes & Stability: Continuous memory optimization, resolving OS-specific crashes (such as Linux Wayland/WebKit edge cases), and refining GPU processing backend auto-selection for seamless performance across all platforms.
• RapidRAW Cloud: Refining the newly launched cloud inpainting service based on early feedback, with a focus on speed and quality for large removals.
• Performance & Algorithm Refinements: Further optimizing processing speed on older GPU architectures, refining Fujifilm X-Trans sensor demosaicing algorithms, and expanding UI responsiveness on mobile/Android devices.

AI Strategy

I've designed RapidRAW's AI architecture with absolute flexibility, privacy, and user choice in mind. Whether you want a purely traditional RAW workflow without any neural networks, fast local automation, powerful self-hosting, or simple cloud convenience, you have four distinct ways to use RapidRAW:

1. Built-in AI Tools (Local & Free)

These features are integrated directly into RapidRAW and run entirely on your local machine, accelerated by your GPU on Windows (DirectML) and macOS (CoreML) with an automatic CPU fallback. They are fast, completely private, free, and require zero setup:

• AI Masking: Instantly detect and mask subjects, skies, foregrounds, and depth (powered by SAM, EfficientSAM3, BiRefNet, and Depth Anything V2).
• AI Text & Face Masks: Select anything by typing what you want, like "car" or "water", or pick individual face and body parts like skin, hair, eyes, or teeth (powered by EfficientSAM3 and MediaPipe).
• AI Denoise: Non-destructive noise reduction for RAW files using the NIND model.
• Spatial Effects: Depth and surface normal estimation for Lens Blur, Relight, and Fog (powered by Depth Anything V2 and MoGe-2).
• Automatic Tagging: The image library is automatically indexed with keywords using a local CLIP model for effortless search.
• Simple Generative Replace: A lightweight inpainting tool (via LaMa) for removing sensor dust, blemishes, and small distractions.

2. Self-Hosted Integration with ComfyUI (Local & Free)

For users with a capable GPU who want maximum control over generative features, RapidRAW connects directly to your local ComfyUI server through the RapidRAW-AI-Connector.

Why this approach? The AI Connector acts as a high-performance middleware that handles image caching, workflow injection, and coordination. Instead of transferring the entire high-resolution image for every edit, the full image is cached once; subsequent generative tweaks transfer only the tiny mask and prompt.

• Full Control: Use your own local hardware and any custom Diffusion model or checkpoint.
• Cost-Free Power: Utilize your existing GPU for heavy generative edits without subscriptions.
• Custom Workflow Selection: Import your own ComfyUI node graphs and workflows into RapidRAW-AI-Connector.

3. Optional Cloud Service (Subscription)

To be clear: core software functionality is never locked behind a paywall. All features remain free if you use built-in tools or choose to self-host.

However, not everyone has access to a powerful GPU or wants the technical overhead of configuring and maintaining a local ComfyUI server. For users who prefer seamless convenience, RapidRAW offers an optional RapidRAW Cloud subscription for $14/month:

• Turnkey Setup: No local server or model weights to download. Just log in and run high-end inpainting/generative tools immediately.
• Support Indie Development: Subscribing is the best way to directly fund continuous development, updates, and maintenance.

4. AI-Free Mode (Local & Distraction-Free)

For purists, contest photographers, or those with ethical or privacy-related objections to AI tools, RapidRAW can function as a purely traditional, non-AI RAW editor with a single toggle in Settings → General:

• Decluttered UI: All AI-related buttons, generative patch menus, AI selection tabs, and AI-driven lens blur tools are hidden from the interface.
• Algorithmic Fallbacks: Features like Denoise automatically fall back to classic algorithmic filters (such as BM3D) and disable neural-based models.
• Non-Destructive Compatibility: Any existing AI masks or edits in previously edited sidecar files (.rrdata) will still render faithfully without breaking your library.

---

Strategy Comparison

Feature	Built-in AI (Free)	Self-Hosted (ComfyUI)	Optional Cloud Service	AI-Free Mode
Cost	Free, included	Free (requires your own GPU)	$14 / month	Free, included
Setup	None	Manual ComfyUI / AI Connector setup	None (Just log in)	Single switch in Settings
Privacy	100% Local execution	100% Local / Self-hosted	Processed via Cloud API	100% Local (Zero AI execution)
Use Case	Everyday workflow acceleration	Maximum control for power users	Maximum convenience	Traditional editing & contest workflows

Initial Development Log

This project began as an intensive sprint to build the core functionality. Here's a summary of the initial progress and key milestones:

• Day 1: June 13th, 2025 - Project inception, basic Tauri setup, and initial brightness/contrast shader implementation.
• Day 2: June 14th - Core architecture refactor, full library support (folder tree, image list), and optimized image loading. Implemented histogram and curve editor support. Added UI themes.
• Day 3: June 15th - Implemented a working crop tool, preset system, and context menus. Enabled auto-saving of edits to sidecar files and auto-thumbnail generation. Refined color adjustments.
• Day 4: June 16th - Initial prototype for local adjustments with masking. Added mask support to presets. Bug-free image preview switching.
• Day 5: June 17th - Major UI overhaul. Created the filmstrip and resizable panel layout. Fixed mask scaling issues and improved the library/welcome screen.
• Day 6: June 18th - Performance tuning. Reduced GPU calls for adjustments, leading to a much smoother cropping and editing experience. Implemented saving of panel UI state.
• Day 7: June 19th - Enhanced library functionality. Added multi-selection and the ability to copy/paste adjustments across multiple images.
• Day 8: June 20th - Implemented initial RAW file support and an EXIF metadata viewer.
• Day 9: June 21st - Added advanced detail adjustments (Clarity, Sharpening, Dehaze, etc.) and film grain. Developed a linear RAW processing pipeline.
• Day 10: June 22nd - Implemented layer stacking for smooth preview transitions. Built a robust export panel with batch export capabilities. Added import/export for presets.
• Day 11: June 23rd - Added full undo/redo functionality integrated with a custom history hook. Improved context menus and completed the settings panel.
• Day 12: June 24th - Implemented image rotation and fixed all mask scaling/alignment issues related to cropping and rotation.
• Day 13: June 25th - Rewrote the mask system to be bitmap-based. Implemented brush and linear gradient tools, with semi-transparent visualization.
• Day 14: June 26th-27th - Final polish. Added universal keyboard shortcuts, full adjustment support for masks, theme management, and final UI/UX improvements. This ReadMe.

Getting Started

You have two options to run RapidRAW:

1. Download the Latest Release (Recommended)

The download page picks the right file for your system. All builds are also on the Releases page.

Windows & macOS:

• Grab the installer or application bundle for your operating system from the download page.

Linux:

• The official Flatpak package supports all Linux distributions and is available on Flathub.
• On Debian-based distributions, install the .deb package from the download page.
• On Arch-based distributions, use the rapidraw-bin package from the AUR.

Android:

• Install the .apk from the download page. Requires a 64-bit ARM device with Android 7.0 or newer.

2. Build from Source

If you want to build the project yourself, you'll need to have Rust and Node.js installed.

# 1. Clone the repository
git clone https://github.com/CyberTimon/RapidRAW.git
cd RapidRAW

# 2. Install frontend dependencies
npm install

# 3. Build and run the application
npm start

To build the release build:

# 1. Make the release build
npm run tauri build

# 2. Run the release build
./src-tauri/target/release/RapidRAW

To build with camera tethering support, install the system dependencies first:

# macOS (Homebrew)
brew install libgphoto2 pkg-config

# Linux (Ubuntu / Debian)
sudo apt-get install -y libgphoto2-dev pkg-config

Then run or build using the tethering feature flag:

# Development mode with Tethering
npm run start:tethering
# or: npm start -- -- --features tethering

# Release build with Tethering
npm run tauri build -- --features tethering

Camera Tethering

RapidRAW includes camera tethering for studio, portrait, and product photography workflows. Connect your camera via USB to control exposure settings, monitor your shot in real time, and automatically ingest files directly into your workspace.

Supported Cameras

Tethering is powered by libgphoto2 and supports over 2,500+ camera models across Canon, Nikon, Sony, Fujifilm, Olympus, Panasonic, and other manufacturers.

• View the Full List of Supported Cameras

Note: Ensure your camera's USB connection mode is set to PC Remote, Tether Shooting, or PTP in the camera settings.

Platform Support & Installation

Tethering is supported on macOS and Linux. Windows and Android are not supported.

The tethering build dynamically links directly against libgphoto2. If these system libraries are not present on a machine, an executable linked against them will fail to launch entirely. To ensure standard RapidRAW builds remain dependency-free and start on any system without extra installation steps, tethering is distributed as a dedicated _tethering release.

libgphoto2 is designed for POSIX environments and requires direct low-level access via libusb. On Windows, connected cameras are claimed by the operating system's native Windows Portable Device (WPD) drivers. Interfacing with libgphoto2 on Windows requires overriding OEM drivers with WinUSB (using tools like Zadig), which breaks standard file transfer and camera utilities. Because of these driver conflicts and platform limitations, Windows is fully unsupported.

Setting up the Tethering Build

To run the _tethering release binaries (or build from source), you must install libgphoto2 on your machine first.

macOS (via Homebrew):

brew install libgphoto2 pkg-config

Linux (Ubuntu / Debian):

sudo apt-get update
sudo apt-get install -y libgphoto2-dev pkg-config

Linux (Arch Linux):

sudo pacman -S libgphoto2 pkgconf

Linux (Fedora):

sudo dnf install libgphoto2-devel pkgconf-pkg-config

After installing the system dependencies, download the corresponding tethering build from the download page, or run/build from source using the tethering feature flag.

Command Line Interface (CLI)

RapidRAW includes a headless export tool for batch processing photos in automated scripts, terminal pipelines, or server environments without opening the GUI:

# Export an entire folder using edits found in .rrdata sidecar files
rapidraw export /path/to/photos --output /path/to/output_dir --format jpeg --quality 90

# Export a single image directly to a specific target file
rapidraw export /path/to/photo.raw --output /path/to/output.png --format png

# Export a true 16-bit TIFF (the TIFF default; use 8 for an RGB8 TIFF)
rapidraw export /path/to/photo.raw --output /path/to/output.tiff --format tiff --tiff-bit-depth 16

# Batch export a folder using a custom adjustments JSON file to override sidecars
rapidraw export /path/to/photos --output /path/to/output_dir --adjustments /path/to/preset.json

Note: By default, headless export automatically detects and applies edits stored in .rrdata sidecar files located alongside your source images. You can override sidecars for all exported images by passing a custom JSON file using the --adjustments flag.

Option	Description	Default
<source>	Path to an image file or directory containing images	(Required)
--output <path>	Target directory or specific output file path	(Required)
--format <fmt>	Output format (jpeg, png, webp, avif, tiff, jxl, cube)	jpeg
--quality <1-100>	Image export quality	90
--tiff-bit-depth <n>	TIFF channel depth (8 or 16)	16
--keep-metadata	Retain EXIF/capture metadata in exported files	false
--adjustments <path>	Path to a custom JSON file containing adjustments to override sidecars	(Auto-detected)

Companion Apps & Ecosystem

The community has created dedicated standalone companion tools designed to integrate directly with RapidRAW's non-destructive editing workflow and .rrdata sidecars:

• Immich: A self-hosted photo and video library. RapidRAW connects directly to your Immich server (2.5 or newer) to browse your albums and timeline, edit photos, and export back to Immich. Your edits are stored with each photo on the server, so they stay in sync across all your computers.
• RapidTimelapse by @crispin81: An LRTimelapse-style keyframe ramping and deflicker utility. Edit key turning-point frames in RapidRAW, mark them with 5 stars, and RapidTimelapse will smoothly calculate and write interpolated adjustments straight to intermediate .rrdata sidecars.
• RapidReady by @on370 (Website): A high-speed digital asset manager (DAM), ingest, and culling tool. It offers zero-latency browsing through instant embedded preview extraction, Difference-of-Gradients focus peaking, GPS geotagging, virtual albums, and bidirectional rating and metadata synchronization with RapidRAW.

Developed a tool, script, or workflow integration for RapidRAW? Please open a pull request to have it listed here!

System Requirements

RapidRAW is built to be lightweight and cross-platform. The minimum (tested) requirements are:

Operating System:

• Windows: Windows 10 or newer
• macOS: macOS 13 (Ventura) or newer
• Linux: Ubuntu 22.04+ or a compatible modern distribution

Hardware Recommendations:

• RAM: 16GB or more is highly recommended. While the application may run on systems with less memory, performance is best with 16GB+ to handle high-resolution RAW files, undo history, and complex layer masking without slowdowns.
• GPU: A dedicated GPU is recommended. RapidRAW relies heavily on GPU acceleration for its processing pipeline. Very old GPU architectures (generally pre-2015) or older integrated graphics may struggle, leading to instability or graphical artifacts.

Common Problems

If the application crashes immediately when you try to start editing a picture, it is often due to the automatic selection of the GPU backend.

1. Open Settings on the Home Screen (Gear icon).
2. Navigate to the Processing tab.
3. Locate the Processing Backend setting.
4. Change it from Auto to a specific backend supported by your OS (e.g., Vulkan, DirectX12, OpenGL, or Metal).
5. Restart the application and try opening the image again. Experiment with different backends if the first one doesn't work.

If RapidRAW crashes on Wayland (e.g. GNOME + NVIDIA), try launching it with:

WEBKIT_DISABLE_DMABUF_RENDERER=1 RapidRAW

or

WEBKIT_DISABLE_COMPOSITING_MODE=1 RapidRAW

This issue is related to WebKit and NVIDIA drivers, not RapidRAW directly. Switching to X11 or using AMD / Intel GPUs may also help.

See #306 for more information.

Contributing

I’m really grateful for any contributions you make to RapidRAW! Whether you’re reporting a bug, suggesting a new feature, or submitting a pull request - your input helps shape the project and makes it better for everyone. Don’t hesitate to open an issue or share your ideas.

Before diving into the code, the Technical Reference is a good starting point. It describes every stage of the processing pipeline, which source files implement it, and how the individual algorithms work.

Image format issues

If your camera’s RAW files aren’t supported, please open an issue here first: rawler issues. Once support is added in rawler, create an issue for RapidRAW so I can update the packages and keep everything in sync.

Contributing RAW samples

New cameras and RAW formats can only be added to rawler when a full set of CC-0 licensed sample files exists. Without samples there is no test case, and without a test case support can't be merged. Files shared via Google Drive or similar services can't be used.

If you want support for your camera or smartphone, please help by uploading a sample set to RawDB as early as possible:

• Shoot a properly exposed daylight scene with many colors (landscape with trees, fruit basket, colorful houses). No people, cars or other private property, and no color checker cards.
• Cover the modes your camera supports (e.g. ISO levels, compression, bit depth, crop and RAW sizes). The required set differs per manufacturer.
• Give your files descriptive names like CANON_5DMK3_ISO_100_SRAW.CR2 instead of IMG_0001.CR2.
• By uploading, you confirm that you are the copyright owner and license the samples under CC-0.

Please read the full sample guidelines before shooting, and link your RawDB upload in the dnglab issue.

Special Thanks

A huge thank you to the following projects and tools that were very important in the development of RapidRAW:

• Google AI Studio: For providing amazing assistance in researching, implementing image processing algorithms and giving an overall speed boost.
• rawler: For the excellent Rust crate that provides the foundation for RAW file processing in this project.
• lensfun: For its invaluable open-source library and comprehensive database for automatic lens correction.
• LaMa: For the powerful & simple image inpainting model, which enables content-aware fill and object removal.
• Segment Anything: For providing the foundation model used for the AI subject detection capabilities.
• EfficientSAM3: For the efficient promptable segmentation model distilled from SAM 3 that powers the AI sky and text masks.
• BiRefNet: For the high-resolution segmentation model that powers the AI foreground detection.
• MediaPipe: For the fast face landmark and segmentation models that power the AI face masks.
• Depth Anything V2: For the powerful monocular depth estimation model that enables the AI depth masking capabilities.
• MoGe-2: For the surface normal estimation model that powers the Relight effect.
• nind-denoise: For providing AI models that power the AI noise reduction capabilities in RapidRAW.

Support the Project

As a young developer balancing this project with an apprenticeship, your support means the world. If RapidRAW has a place in your workflow, a donation buys me more time to work on it, new gear to test it with, and days outdoors with my camera, which is where most new ideas for RapidRAW start.

• Ko-fi: Donate on Ko-fi
• Crypto:
  • BTC: 36yHjo2dkBwQ63p3YwtqoYAohoZhhUTkCJ (min. 0.0001)
  • ETH: 0x597e6bdb97f3d0f1602b5efc8f3b7beb21eaf74a (min. 0.005)
  • SOL: CkXM3C777S8iJX9h3MGSfwGxb85Yx7GHmynQUFSbZXUL (min. 0.01)

License & Philosophy

This project is licensed under the GNU Affero General Public License v3.0 (AGPL-3.0). I chose this license to ensure that RapidRAW and any of its derivatives will always remain open-source and free for the community. It protects the project from being used in closed-source commercial software, ensuring that improvements benefit everyone.

See the LICENSE file for more details.