RapidRAW v1.6.7 released: major performance boost, Immich integration, smart AI masks, GPU-accelerated AI & more
Timon has released v1.6.7. The headline is a rewritten image pipeline that
makes the app faster and lighter on RAM, alongside Immich integration, GPU
acceleration for AI models, new AI masks and a long list of image quality
rewrites. This is a pretty big one so check out the official release notes for full context with screenshots: RapidRAW v1.6.7 on GitHub
Some existing edits may look different
Several core algorithms were rewritten, so older edits can shift after updating:
- RAW colors: tungsten and artificial light photos may change, and very
saturated colors are now compressed instead of clipped.
- Dehaze and Whites: images using these may need readjusting.
- Noise reduction and sharpening: may look cleaner or sharper.
- Highlights: clipped RAW highlights are rebuilt with more natural color.
If you're mid-project on something important, finish it on your current version
first.
Performance
Faster thumbnails, previews, exports and batch actions, optimized GPU shaders
and lower memory use. Two new Advanced settings let you control export
concurrency and how much RAM a batch export may use.
New features
Immich integration
Browse your Immich albums and timeline and edit photos as if they were local
files. Edits are stored on the Immich server and synced, you can export straight
to Immich, and the API key is kept in your system's credential store. Requires
Immich 2.5 or newer. Thanks to contributor cl1x.
AI and masking
- GPU acceleration for AI models: DirectML on Windows, CoreML on macOS,
with automatic CPU fallback. Can be turned off in settings.
- Smart AI masks in copy & paste and presets: AI masks now regenerate for
each image instead of being pasted as-is. Lens Blur, Fog, Relight and AI
Denoise are now available as copy & paste groups.
- New sky and foreground models: EfficientSAM3 for sky and BiRefNet for
foreground, both with much finer edges. They download on first use.
- AI Text Mask: type the name of an object and get a mask for it.
- Face Segmentation: select facial regions such as skin, hair, eyes, lips
and teeth.
Denoise, now fully non-destructive
AI Denoise runs on the GPU on Windows and macOS and can be pasted across photos
without creating extra files. The old TIFF-saving denoise modal is gone. For
AI-Free Mode there's a new BM3D-based Classic Denoise in the Details panel.
Non-destructive negative conversion
Film negative conversion now lives in the Details panel with adjustable color
balance, print exposure and print grade. Batch conversion is in the right-click
menu. The old modal is removed.
Job Queue
Replaces the old AI queue and tracks exports, imports, file transfers, model
downloads and AI tasks in one place. Everything can be cancelled, and AI jobs
can be paused, resumed or retried.
Image quality
- Color-preserving highlight reconstruction working directly on RAW data.
- More accurate RAW colors with dual-illuminant camera profiles and gamut
compression.
- Rewritten noise reduction and sharpening with better detail and smoother
curved edges.
- Rewritten Dehaze that estimates haze color and removes it with an
edge-aware filter. Sliding left still adds haze.
Workflow
- New before/after compare modes: split left/right, split top/bottom and
side by side, keeping zoom and position.
- Auto Adjust levels: Subtle, Normal and Strong from the right-click menu.
Auto now only touches exposure, not Brightness.
- Storage page: see disk use by category, clear each one, and set size
limits with automatic cleanup.
Fixes and smaller improvements
- Fast inpainting keeps more surrounding texture.
- Better AI mask edges, and fixed banding on face, hair and skin masks.
- Tethering: cameras that reuse file names no longer show the previous shot.
- Thumbnails refresh when RAW processing settings change.
- Stronger Whites slider reaching further into the midtones.
- Fixed a jump in the Highlights slider when leaving zero.
- Fixed loading from network drives without file locking.
- Up/down arrow keys move between rows in the library grid.
- An undo toast when a rating or flag removes an image from the current filter.
- General stability and performance fixes.
Downloads
Standard builds for Windows, macOS (Apple Silicon and Intel), Ubuntu 22.04 and
24.04, and Android. Separate _tethering builds for macOS and Ubuntu (these
need libgphoto2 installed first). ARM builds, RPM and AppImage are in the full
asset list.
RapidRAW v1.6.7 on GitHub
A note on installation warnings
RapidRAW is not yet code-signed, so your OS may complain on first launch:
- macOS: if macOS reports the app as damaged or corrupted, it is not
broken. Remove the quarantine flag after installation:
xattr -dr com.apple.quarantine /Applications/RapidRAW.app
- Windows: SmartScreen may warn you. Click More info → Run anyway.
31 points · 18 comments
RapidRAW now integrates with Immich
Hi everyone,
I'm the developer of RapidRAW, a RAW editor with over 10k stars on GitHub, and the latest release adds Immich integration.
If you’ve been looking for a way to edit your RAWs directly from your Immich library, this might be worth a look. RapidRAW is fast, runs fully local, and covers the usual stuff like masking, color grading, lens correction and denoise.
You connect RapidRAW to your server and can then browse your albums and timeline right in the library and edit any photo like a local file. Originals are only downloaded when you open or export them, so you don't need a local copy of your library.
Edits are stored with each photo on your Immich server (this needs Immich 2.5 or newer), so they look the same on every computer and survive a cleared cache. Your originals are never changed. You can also export straight back to Immich, where the export is stacked on top of its original.
Big thanks to cl1x, who built the integration.
preview.redd.it
Release and downloads: github.com/CyberTimon/RapidRAW/releases/latest
Website: getrapidraw.com
It's the first release with Immich support, so I'd love to hear how it works with your setup.
Timon
132 points · 16 comments
RapidRAWの開発者による2.5Dリライティング実装の告知とディスカッションが行われているRedditスレッド
reddit.com
Rust / Tauri / WGPUベースで開発されたオープンソースの写真編集ソフトウェア「RapidRAW」のGitHubリポジトリ
CyberTimon/RapidRAW
272 views
I implemented 2.5D relighting using depth and normal maps (alternative to diffusion relighting) to my open source photo editor
Hey guys,
For a long time I tried using different Stable Diffusion relighting setups and models, but I always ran into the same issues on high-res renders and photos. Because they go through VAE encoding and img2img diffusion, you almost always get slight identity loss, changed details, or compression artifacts.
In the latest update of my open source photo editor (RapidRAW), I implemented a different approach: 2.5D relighting using depth maps and surface normals.
Instead of running the image through a generative model, it estimates depth and surface normals locally, and then calculates the lighting directly on top of your original pixels.
Because it does not regenerate the image:
- Zero VAE compression or blur
- Original details and face identity stay 100% identical
- You can move point lights, spot lights, and ray-marched shadows in real time
- It works on full high resolution images without running out of VRAM
The downside compared to diffusion is that it won't hallucinate completely new bounced reflections or change complex backgrounds. But if you just need to fix lighting, add a spot light, or cast shadows on an existing generation without changing any pixels, it works really well.
The app is completely free, local, and open source (Rust/Tauri/WGPU). Thought people here who struggle with diffusion relighting changing their images might find it useful:
GitHub: CyberTimon/RapidRAW
preview.redd.it
preview.redd.it
Let me know what you think or if you have any feedback.
Timon
11 points · 4 comments
RapidRAW v1.6.6 released: Spatial Effects & 2.5D Relighting, non-destructive AI Denoise, Auto Straighten & more
Timon has released v1.6.6. The focus this time is spatial creative tools
and automation, plus some important changes to how RAW exposure is handled.
New features
Spatial Effects and 2.5D Relighting
RapidRAW now estimates a depth map and surface normals from your image, which
enables depth-aware effects in real time:
- Relighting with point, spot and directional lights that respond to the
surfaces in the scene.
- Ray-traced shadows cast against the reconstructed depth.
- Volumetric and ground fog that follows scene height and distance.
Non-destructive AI Denoise
The denoiser now lives in the Details panel as a normal tool, with sliders for
reduction strength and fine detail recovery. No more context-menu action
producing a large separate TIFF. GPU acceleration and non-destructive batch
denoising are still in development.
Auto Straighten
Analyzes horizons, architectural edges and contours to work out the rotation
and reframe the crop. Use Shift+S in the Crop panel, the right-click menu,
or run it in batch from the library grid. Detection accuracy is still being
refined, so expect the odd miss.
Custom center point for Vignette and Centré
Drag an on-canvas pin to move the center of the effect, with a live contour
overlay. Double-click the pin to reset it.
Individual tool visibility toggles
Every tool and sub-section has its own on/off toggle, so you can compare a
single adjustment like grain, sharpening or color grading without switching off
the rest of your edit.
RAW baseline exposure and highlight preservation compensation
- A configurable baseline exposure bias, +0.7 EV by default, plus support
for embedded DNG baseline tags.
- Restores the exposure held back by in-camera highlight protection modes on
Fujifilm, Canon, Nikon, Olympus and Pentax.
- Both are adjustable in Preferences and tracked per image, so photos you've
already edited are not affected.
Core improvements
- DPI setting in the export dialog.
- Keywords, ratings and color labels can be written as standard XMP metadata.
- Touch multi-selection on Android for rating and batch editing.
- New {original_filename_number} batch rename token to keep in-camera frame
numbers.
- Updated Lensfun database for broader lens correction support.
- .rrdata sidecars are no longer written for images with only default settings.
- Staged preview caching to reduce lag and flicker when dragging sliders.
- General stability and performance fixes.
Downloads
Standard builds for Windows, macOS (Apple Silicon and Intel), Ubuntu 22.04 and
24.04, and Android. Separate _tethering builds for macOS and Ubuntu (these
need libgphoto2 installed first). ARM builds, RPM and AppImage are in the full
asset list.
RapidRAW v1.6.6 on GitHub
A note on installation warnings
RapidRAW is not yet code-signed, so your OS may complain on first launch:
- macOS: if macOS reports the app as damaged or corrupted, it is not
broken. Remove the quarantine flag after installation:
xattr -dr com.apple.quarantine /Applications/RapidRAW.app
- Windows: SmartScreen may warn you. Click More info → Run anyway.
29 points · 5 comments
@dhh cybertimon/rapidraw if darktable seems like dark magic
1 like · 420 views
@htormey there's this one too cybertimon/rapidraw
1 like · 49 views
RapidRAW v1.6.5 released: Kelvin white balance, edge-aware Guided Filter, Pick & Reject flags, and RapidRAW Cloud
Timon has released v1.6.5, focused on image quality and color science, with
a big batch of workflow improvements on top.
Read before upgrading
White balance handling has changed significantly with the new Kelvin engine.
Existing edits may look slightly different in v1.6.5, possibly shifting a
little magenta or green. If you're in the middle of a client shoot or important
project, finish and export it on your current version first, then upgrade.
New features
Kelvin white balance with area sampling
RapidRAW now reads the white balance your camera recorded in the RAW file and
lets you work in true Kelvin temperatures. The eyedropper has been redesigned:
a floating card shows live color readings as you move, and you can drag a box
over a neutral grey or white area to average out noise and calculate the
correct balance.
Guided Filter for edge-aware adjustments
Clarity, structure, dehaze, shadows and highlights now run through an
edge-aware filter instead of a standard blur. Result: no more halos around
high-contrast edges, and textures like skin, fabric and foliage stay crisp.
Rewritten Whites slider
Instead of a flat multiplier that blew out highlights, Whites now uses a smooth
targeted curve and reinjects fine detail, so clouds, snow and white clothing
gain brilliance without turning into flat patches.
Pick and Reject flags
A dedicated culling workflow. Mark photos as Picked or Rejected with keyboard
shortcuts, badges or bottom bar toggles. Rejected photos dim in the library and
can be filtered out or deleted in one go.
Apple RAW 9 decoding and denoising (macOS)
Mac users can use Apple's RAW 9 engine, which demosaics and denoises in a single
step on the Neural Engine. Enable it as the default RAW developer in processing
settings, or pick it in the Denoise menu. Unsupported cameras fall back to the
built-in processor automatically.
AI-Free Mode
One switch hides all generative buttons, AI masks and neural tools, and reverts
features like noise reduction to classic algorithms. Useful for contest rules or
if you just prefer a traditional workspace. Existing edits still display normally.
RapidRAW Cloud (optional)
A new optional paid service ($14/month) for generative object removal on remote
GPUs, aimed at people without a capable graphics card or who don't want to set
up a local server. Only the small patch you brush over is sent and it's deleted
after processing, so full RAW files stay on your machine.
The editor itself, RAW engine, masking and built-in spot removal remain free and
local, and connecting your own local AI server still works at no cost. It's an
early rollout, so feedback on speed and quality is welcome.
More info: getrapidraw.com
Reorderable and collapsible panels
Drag to reorder entire sections and individual tools, hide ones you rarely use,
and collapse the rest. Your layout is remembered.
Savable custom aspect ratios
Save any custom crop ratio as a preset next to the standard ones.
Core improvements
- Tone curves: fine adjust mode, plus inactive channels shown as colored
overlays with reference markers.
- Crop: hold Ctrl/Cmd for center-anchored resizing, new center mark guide.
- Export: borders, aspect ratio padding, embedded sRGB ICC profiles, refreshed
export presets.
- AI tasks now show real-time progress and can be cancelled.
- Better lens EXIF detection and refined RAW magenta highlight correction.
- Dutch added to the language list.
- General stability and performance fixes.
Big thanks to contributor lalibertemarc, who kept development moving while
CyberTimon was away.
Downloads
Standard builds for Windows, macOS (Apple Silicon and Intel), Ubuntu 22.04 and
24.04, and Android. Separate _tethering builds for macOS and Ubuntu (these
need libgphoto2 installed first). ARM builds, RPM and AppImage are in the full
asset list.
RapidRAW v1.6.5 on GitHub
A note on installation warnings
RapidRAW is not yet code-signed, so your OS may complain on first launch:
- macOS: if macOS reports the app as damaged or corrupted, it is not
broken. Remove the quarantine flag after installation:
xattr -dr com.apple.quarantine /Applications/RapidRAW.app
- Windows: SmartScreen may warn you. Click More info → Run anyway.
20 points · 7 comments
Mentions
RapidRAW v1.6.7 released: major performance boost, Immich integration, smart AI masks, GPU-accelerated AI & more Timon has released v1.6.7. The headline is a rewritten image pipeline that makes the app faster and lighter on RAM, alongside Immich integration, GPU acceleration for AI models, new AI masks and a long list of image quality rewrites. This is a pretty big one so check out the official release notes for full context with screenshots: RapidRAW v1.6.7 on GitHub Some existing edits may look different Several core algorithms were rewritten, so older edits can shift after updating: - RAW colors: tungsten and artificial light photos may change, and very saturated colors are now compressed instead of clipped. - Dehaze and Whites: images using these may need readjusting. - Noise reduction and sharpening: may look cleaner or sharper. - Highlights: clipped RAW highlights are rebuilt with more natural color. If you're mid-project on something important, finish it on your current version first. Performance Faster thumbnails, previews, exports and batch actions, optimized GPU shaders and lower memory use. Two new Advanced settings let you control export concurrency and how much RAM a batch export may use. New features Immich integration Browse your Immich albums and timeline and edit photos as if they were local files. Edits are stored on the Immich server and synced, you can export straight to Immich, and the API key is kept in your system's credential store. Requires Immich 2.5 or newer. Thanks to contributor cl1x. AI and masking - GPU acceleration for AI models: DirectML on Windows, CoreML on macOS, with automatic CPU fallback. Can be turned off in settings. - Smart AI masks in copy & paste and presets: AI masks now regenerate for each image instead of being pasted as-is. Lens Blur, Fog, Relight and AI Denoise are now available as copy & paste groups. - New sky and foreground models: EfficientSAM3 for sky and BiRefNet for foreground, both with much finer edges. They download on first use. - AI Text Mask: type the name of an object and get a mask for it. - Face Segmentation: select facial regions such as skin, hair, eyes, lips and teeth. Denoise, now fully non-destructive AI Denoise runs on the GPU on Windows and macOS and can be pasted across photos without creating extra files. The old TIFF-saving denoise modal is gone. For AI-Free Mode there's a new BM3D-based Classic Denoise in the Details panel. Non-destructive negative conversion Film negative conversion now lives in the Details panel with adjustable color balance, print exposure and print grade. Batch conversion is in the right-click menu. The old modal is removed. Job Queue Replaces the old AI queue and tracks exports, imports, file transfers, model downloads and AI tasks in one place. Everything can be cancelled, and AI jobs can be paused, resumed or retried. Image quality - Color-preserving highlight reconstruction working directly on RAW data. - More accurate RAW colors with dual-illuminant camera profiles and gamut compression. - Rewritten noise reduction and sharpening with better detail and smoother curved edges. - Rewritten Dehaze that estimates haze color and removes it with an edge-aware filter. Sliding left still adds haze. Workflow - New before/after compare modes: split left/right, split top/bottom and side by side, keeping zoom and position. - Auto Adjust levels: Subtle, Normal and Strong from the right-click menu. Auto now only touches exposure, not Brightness. - Storage page: see disk use by category, clear each one, and set size limits with automatic cleanup. Fixes and smaller improvements - Fast inpainting keeps more surrounding texture. - Better AI mask edges, and fixed banding on face, hair and skin masks. - Tethering: cameras that reuse file names no longer show the previous shot. - Thumbnails refresh when RAW processing settings change. - Stronger Whites slider reaching further into the midtones. - Fixed a jump in the Highlights slider when leaving zero. - Fixed loading from network drives without file locking. - Up/down arrow keys move between rows in the library grid. - An undo toast when a rating or flag removes an image from the current filter. - General stability and performance fixes. Downloads Standard builds for Windows, macOS (Apple Silicon and Intel), Ubuntu 22.04 and 24.04, and Android. Separate _tethering builds for macOS and Ubuntu (these need libgphoto2 installed first). ARM builds, RPM and AppImage are in the full asset list. RapidRAW v1.6.7 on GitHub A note on installation warnings RapidRAW is not yet code-signed, so your OS may complain on first launch: - macOS: if macOS reports the app as damaged or corrupted, it is not broken. Remove the quarantine flag after installation: xattr -dr com.apple.quarantine /Applications/RapidRAW.app - Windows: SmartScreen may warn you. Click More info → Run anyway.
RapidRAW now integrates with Immich Hi everyone, I'm the developer of RapidRAW, a RAW editor with over 10k stars on GitHub, and the latest release adds Immich integration. If you’ve been looking for a way to edit your RAWs directly from your Immich library, this might be worth a look. RapidRAW is fast, runs fully local, and covers the usual stuff like masking, color grading, lens correction and denoise. You connect RapidRAW to your server and can then browse your albums and timeline right in the library and edit any photo like a local file. Originals are only downloaded when you open or export them, so you don't need a local copy of your library. Edits are stored with each photo on your Immich server (this needs Immich 2.5 or newer), so they look the same on every computer and survive a cleared cache. Your originals are never changed. You can also export straight back to Immich, where the export is stacked on top of its original. Big thanks to cl1x, who built the integration. preview.redd.it Release and downloads: github.com/CyberTimon/RapidRAW/releases/latest Website: getrapidraw.com It's the first release with Immich support, so I'd love to hear how it works with your setup. Timon
RapidRAWの開発者による2.5Dリライティング実装の告知とディスカッションが行われているRedditスレッド reddit.com Rust / Tauri / WGPUベースで開発されたオープンソースの写真編集ソフトウェア「RapidRAW」のGitHubリポジトリ CyberTimon/RapidRAW
I implemented 2.5D relighting using depth and normal maps (alternative to diffusion relighting) to my open source photo editor Hey guys, For a long time I tried using different Stable Diffusion relighting setups and models, but I always ran into the same issues on high-res renders and photos. Because they go through VAE encoding and img2img diffusion, you almost always get slight identity loss, changed details, or compression artifacts. In the latest update of my open source photo editor (RapidRAW), I implemented a different approach: 2.5D relighting using depth maps and surface normals. Instead of running the image through a generative model, it estimates depth and surface normals locally, and then calculates the lighting directly on top of your original pixels. Because it does not regenerate the image: - Zero VAE compression or blur - Original details and face identity stay 100% identical - You can move point lights, spot lights, and ray-marched shadows in real time - It works on full high resolution images without running out of VRAM The downside compared to diffusion is that it won't hallucinate completely new bounced reflections or change complex backgrounds. But if you just need to fix lighting, add a spot light, or cast shadows on an existing generation without changing any pixels, it works really well. The app is completely free, local, and open source (Rust/Tauri/WGPU). Thought people here who struggle with diffusion relighting changing their images might find it useful: GitHub: CyberTimon/RapidRAW preview.redd.it preview.redd.it Let me know what you think or if you have any feedback. Timon
RapidRAW v1.6.6 released: Spatial Effects & 2.5D Relighting, non-destructive AI Denoise, Auto Straighten & more Timon has released v1.6.6. The focus this time is spatial creative tools and automation, plus some important changes to how RAW exposure is handled. New features Spatial Effects and 2.5D Relighting RapidRAW now estimates a depth map and surface normals from your image, which enables depth-aware effects in real time: - Relighting with point, spot and directional lights that respond to the surfaces in the scene. - Ray-traced shadows cast against the reconstructed depth. - Volumetric and ground fog that follows scene height and distance. Non-destructive AI Denoise The denoiser now lives in the Details panel as a normal tool, with sliders for reduction strength and fine detail recovery. No more context-menu action producing a large separate TIFF. GPU acceleration and non-destructive batch denoising are still in development. Auto Straighten Analyzes horizons, architectural edges and contours to work out the rotation and reframe the crop. Use Shift+S in the Crop panel, the right-click menu, or run it in batch from the library grid. Detection accuracy is still being refined, so expect the odd miss. Custom center point for Vignette and Centré Drag an on-canvas pin to move the center of the effect, with a live contour overlay. Double-click the pin to reset it. Individual tool visibility toggles Every tool and sub-section has its own on/off toggle, so you can compare a single adjustment like grain, sharpening or color grading without switching off the rest of your edit. RAW baseline exposure and highlight preservation compensation - A configurable baseline exposure bias, +0.7 EV by default, plus support for embedded DNG baseline tags. - Restores the exposure held back by in-camera highlight protection modes on Fujifilm, Canon, Nikon, Olympus and Pentax. - Both are adjustable in Preferences and tracked per image, so photos you've already edited are not affected. Core improvements - DPI setting in the export dialog. - Keywords, ratings and color labels can be written as standard XMP metadata. - Touch multi-selection on Android for rating and batch editing. - New {original_filename_number} batch rename token to keep in-camera frame numbers. - Updated Lensfun database for broader lens correction support. - .rrdata sidecars are no longer written for images with only default settings. - Staged preview caching to reduce lag and flicker when dragging sliders. - General stability and performance fixes. Downloads Standard builds for Windows, macOS (Apple Silicon and Intel), Ubuntu 22.04 and 24.04, and Android. Separate _tethering builds for macOS and Ubuntu (these need libgphoto2 installed first). ARM builds, RPM and AppImage are in the full asset list. RapidRAW v1.6.6 on GitHub A note on installation warnings RapidRAW is not yet code-signed, so your OS may complain on first launch: - macOS: if macOS reports the app as damaged or corrupted, it is not broken. Remove the quarantine flag after installation: xattr -dr com.apple.quarantine /Applications/RapidRAW.app - Windows: SmartScreen may warn you. Click More info → Run anyway.
@dhh cybertimon/rapidraw if darktable seems like dark magic
@htormey there's this one too cybertimon/rapidraw
RapidRAW v1.6.5 released: Kelvin white balance, edge-aware Guided Filter, Pick & Reject flags, and RapidRAW Cloud Timon has released v1.6.5, focused on image quality and color science, with a big batch of workflow improvements on top. Read before upgrading White balance handling has changed significantly with the new Kelvin engine. Existing edits may look slightly different in v1.6.5, possibly shifting a little magenta or green. If you're in the middle of a client shoot or important project, finish and export it on your current version first, then upgrade. New features Kelvin white balance with area sampling RapidRAW now reads the white balance your camera recorded in the RAW file and lets you work in true Kelvin temperatures. The eyedropper has been redesigned: a floating card shows live color readings as you move, and you can drag a box over a neutral grey or white area to average out noise and calculate the correct balance. Guided Filter for edge-aware adjustments Clarity, structure, dehaze, shadows and highlights now run through an edge-aware filter instead of a standard blur. Result: no more halos around high-contrast edges, and textures like skin, fabric and foliage stay crisp. Rewritten Whites slider Instead of a flat multiplier that blew out highlights, Whites now uses a smooth targeted curve and reinjects fine detail, so clouds, snow and white clothing gain brilliance without turning into flat patches. Pick and Reject flags A dedicated culling workflow. Mark photos as Picked or Rejected with keyboard shortcuts, badges or bottom bar toggles. Rejected photos dim in the library and can be filtered out or deleted in one go. Apple RAW 9 decoding and denoising (macOS) Mac users can use Apple's RAW 9 engine, which demosaics and denoises in a single step on the Neural Engine. Enable it as the default RAW developer in processing settings, or pick it in the Denoise menu. Unsupported cameras fall back to the built-in processor automatically. AI-Free Mode One switch hides all generative buttons, AI masks and neural tools, and reverts features like noise reduction to classic algorithms. Useful for contest rules or if you just prefer a traditional workspace. Existing edits still display normally. RapidRAW Cloud (optional) A new optional paid service ($14/month) for generative object removal on remote GPUs, aimed at people without a capable graphics card or who don't want to set up a local server. Only the small patch you brush over is sent and it's deleted after processing, so full RAW files stay on your machine. The editor itself, RAW engine, masking and built-in spot removal remain free and local, and connecting your own local AI server still works at no cost. It's an early rollout, so feedback on speed and quality is welcome. More info: getrapidraw.com Reorderable and collapsible panels Drag to reorder entire sections and individual tools, hide ones you rarely use, and collapse the rest. Your layout is remembered. Savable custom aspect ratios Save any custom crop ratio as a preset next to the standard ones. Core improvements - Tone curves: fine adjust mode, plus inactive channels shown as colored overlays with reference markers. - Crop: hold Ctrl/Cmd for center-anchored resizing, new center mark guide. - Export: borders, aspect ratio padding, embedded sRGB ICC profiles, refreshed export presets. - AI tasks now show real-time progress and can be cancelled. - Better lens EXIF detection and refined RAW magenta highlight correction. - Dutch added to the language list. - General stability and performance fixes. Big thanks to contributor lalibertemarc, who kept development moving while CyberTimon was away. Downloads Standard builds for Windows, macOS (Apple Silicon and Intel), Ubuntu 22.04 and 24.04, and Android. Separate _tethering builds for macOS and Ubuntu (these need libgphoto2 installed first). ARM builds, RPM and AppImage are in the full asset list. RapidRAW v1.6.5 on GitHub A note on installation warnings RapidRAW is not yet code-signed, so your OS may complain on first launch: - macOS: if macOS reports the app as damaged or corrupted, it is not broken. Remove the quarantine flag after installation: xattr -dr com.apple.quarantine /Applications/RapidRAW.app - Windows: SmartScreen may warn you. Click More info → Run anyway.