OmaRAW 0.1.0-beta.10

Beta 10 adds luminosity masks and easier stacks, lets you remove a folder from the catalog, and fixes installation on Omarchy edge. It is for Omarchy x86_64 and Omarchy Mac ARM64.

New

• Luminosity masks (#18). In Develop → Masks, choose Luminosity, then Highlights, Midtones, Shadows or Pick a Tone…. The mask covers that range of tones across the whole picture with no drawing. From, To, Softness and Invert refine it.
• Stacks are easier to work with (#19, #21).
  • The filmstrip and grid show how many photos a stack holds, and number each photo in an open stack, 1/5 to 5/5. The filmstrip outlines a stack as one group.
  • The photo you are on when you press Ctrl+G becomes the stack's top.
  • Shift+Left/Right extend the selection.
  • Auto-Stack by Capture Time (Ctrl+Alt+G) works on the selection. The Import panel can stack bursts as they arrive (Stack bursts).
• Remove a folder from the catalog (#20). Right-click a folder and choose Remove from Catalog…. Its photos leave the catalog; the files stay on disk.

Fixed

• Omarchy edge (#22). Beta 8 and 9 installed on systems with OpenEXR 3.5 but could not start. This release adds a second x86_64 package built against current Arch, and the installer picks the one that matches your system. If a later system update moves a library OmaRAW uses, the update still completes and pacman shows how to install the matching OmaRAW build.

Install

Follow the installation instructions, or download your package and SHA256SUMS below:

• …-x86_64.pkg.tar.zst is for Omarchy (stable).
• …-x86_64-openexr3.5.pkg.tar.zst is for systems with OpenEXR 3.5, such as Omarchy edge. pacman -Q openexr shows yours.
• ARM packages are for Omarchy Mac on Asahi Linux.

Validation

All three native packages passed offline import, editing and GUI startup, all seven fully decoded export formats, all five bundled model loads and editable AI selection, with standalone darktable installed. The OpenEXR 3.5 package was built and validated against current Arch (OpenEXR 3.5.2). ARM binaries passed the 16 KiB load-alignment checks. Each package was installed by pacman in a clean root, where its new library check ran after installation and found every library present.

Locally, 73 focused Qt regression checks and 51 Python checks passed. They cover luminosity masks, stacks, folder removal, imports, the installer's package selection, AI workers and RAW denoise with the real models. The new library check and pacman hook were also run against Beta 9 on current Arch, where they named exactly the two missing OpenEXR 3.4 libraries without failing the update. The downloaded x86 package passed isolated offline AI, denoise and export checks. All three complete-source archives match the release commit and their pinned inputs.

Known limitations

The broader local suite skipped 35 optional cases that need additional fixtures, runtimes or hardware. Two pixel-exact reset and preset checks currently fail on the build workstation for this release and for unchanged Beta 9 alike; this is being investigated separately. No new hands-on Mac test is claimed for this release.

Source and checksums

SHA256SUMS covers all three binary packages and their matching complete-source archives. The ARM source archive includes -aarch64 and the OpenEXR 3.5 one -openexr3.5 in its filename. Retain the included third-party notices when redistributing.

OmaRAW 0.1.0-beta.9

Beta 9 is a maintenance release for Omarchy x86_64 and Omarchy Mac ARM64.

Reopening Keyboard shortcuts now clears the previous search text consistently with the displayed results.

Install

Follow the installation instructions, or download your architecture's package and SHA256SUMS below. ARM packages are for Omarchy Mac on Asahi Linux.

Validation

Both native packages passed offline import, editing and GUI startup, all seven fully decoded export formats, all five bundled model loads and editable AI selection. Validation also confirmed that OmaRAW loads its private processing engine with standalone darktable installed. ARM binaries passed the 16 KiB load-alignment checks.

All 20 final local Qt regression results and 47 Python checks passed, covering shortcut interactions, catalog and viewer behavior, installer selection, AI workers and RAW denoise. The downloaded x86 package additionally passed isolated offline AI selection, removal, RAW denoise and export checks. Both complete-source archives match the release commit and their pinned inputs.

Known limitations

The broader local suite skipped 37 optional cases requiring additional fixtures, runtimes or hardware. A fresh Mac Wayland/Vulkan desktop check remains pending; hosted runner validation does not exercise hardware GPU rendering. No new hands-on Mac test is claimed for this release.

Texture-heavy exports and AI RAW denoise can still take time. Performance work is separate from this release.

Source and checksums

SHA256SUMS covers both binary packages and both matching complete-source archives. The ARM source archive includes -aarch64 in its filename. Retain the included third-party notices when redistributing.

OmaRAW 0.1.0 Beta 8 — Omarchy x86 and ARM64

Beta 8 fixes installation and everyday editing issues on Omarchy x86_64 and Omarchy Mac ARM64.

• The x86 package now builds against Omarchy's stable repositories. Each architecture records its actual OpenEXR/OpenJPH requirements, and the installer checks dependencies before installation (#3).
• OmaRAW reliably loads its private processing engine when standalone darktable is also installed (#2).
• Culling keeps your position when picks or rejects leave a filtered view, without skipping the next photograph (#14, #15).
• Sliders and number fields accept scroll events from Wayland touchpad devices (#5).
• One searchable keyboard shortcut window replaces the overflowing cheat sheet; question-mark activation, search focus and Escape dismissal are fixed (#9, #10, #12, #13).
• Help opens with larger text, adjustable saved text size and an expanded reading view (#11).
• The welcome animation uses software decoding by default to avoid unreliable hardware video readback on Asahi. Splash and About show when the interface uses software rendering (#7).
• Command-line help lists all headless commands (#8).

Install

Use an up-to-date Omarchy installation and follow the installation instructions, or download your architecture's package and SHA256SUMS below. ARM packages are for Omarchy Mac on Asahi Linux. On an Asahi Mac, install vulkan-asahi for GPU acceleration if it is missing.

Validation

Both native packages passed offline import, editing and GUI startup, all seven fully decoded export formats, all five bundled model loads and editable AI selection. Validation also confirmed that OmaRAW loads its private engine with standalone darktable installed. ARM binaries passed the 16 KiB load-alignment checks.

Local regression checks passed (14 results), as did all seven installer tests. The downloaded x86 package additionally passed offline AI selection, removal and RAW denoise checks. The software welcome animation was tested natively on the M2 Mac using the final launch sources; a fresh Mac Wayland/Vulkan desktop check was unavailable.

Known performance limitations

Texture-heavy exports and AI RAW denoise can still take time. Performance work is separate from this release.

Source and checksums

SHA256SUMS covers both binary packages and both matching complete-source archives. The ARM source archive includes -aarch64 in its filename. Retain the included third-party notices when redistributing.

OmaRAW 0.1.0 Beta 7 — x86 and ARM64

Beta 7 adds ARM64 packages for Arch Linux ARM and Omarchy Mac on Asahi Linux, alongside the existing x86_64 package. Both architectures are built from the same source revision and include the offline AI models and runtimes.

• The installer chooses the correct architecture and verifies its checksum.
• ARM binaries support Asahi's 16 KiB pages. AI workers discover the Apple Vulkan GPU automatically.
• Native GitHub runners build and validate both packages. Each architecture has a matching complete-source bundle.

Install

Use an up-to-date Omarchy or Arch Linux installation with OpenEXR 3.5 or later. Follow the installation instructions, or download the package matching your machine plus SHA256SUMS below. On an Asahi Mac, install vulkan-asahi for GPU acceleration if it is missing. These are Linux packages, not macOS applications.

Validation

Both actual packages passed offline import, editing, export and GUI checks, with all five bundled AI models checked. Hardware testing of the ARM candidate on an M2 Mac covered the Vulkan interface, AI acceleration and 13 full-resolution exports across all seven formats. The Mac build also received hands-on testing.

Known performance limitations

Texture-heavy exports can be slow on both x86 and ARM; Texture optimization is separate follow-up work. The successful full-resolution batch test excluded Texture. AI RAW denoise works on the tested M2 Mac but can also be slow, depending on the photo and processing backend.

Earlier x86 releases, including Beta 6, remain available. Beta 7 is the first published ARM package.

Source and checksums

SHA256SUMS covers both binary packages and both complete-source archives. The x86 source archive retains its existing filename pattern; the ARM source archive includes -aarch64. Retain the included third-party notices when redistributing.

OmaRAW 0.1.0 Beta 6

Automatic memory-pressure handling and full-size PSD export improvements.

• Large processing buffers stay in RAM while there is room. When memory is under pressure, OmaRAW automatically backs eligible buffers with temporary disk files during editing, previews and exports.
• Temporary working files share the configured disk-cache allowance with browsing previews. Active files are protected during cleanup and abandoned files are reclaimed after a restart. Disk reservations are checked before processing.
• Preferences → Performance now describes the shared disk cache and includes active working data in its usage readout. Existing memory and disk-cache settings continue to apply; no separate switch is needed.
• Full-size PSD export streams its intermediate TIFF in rows, avoiding the extra full-image allocation that prevented some large 16-bit exports.

Disk backing provides reclaimable storage; it is not a general speed improvement or a hard limit on all application RAM. GPU memory, AI workers and some processing scratch allocations remain separate. Accessing disk-backed buffers can take longer than keeping them in RAM.

Development validation included 13 targeted checks and 33 fully decoded, full-size exports from edited 45-megapixel Canon RAW test images across all seven formats. The final package passed separate checks for import/edit/export, interface startup, offline AI, bundled resources and dependencies. Small CPU/GPU pixel differences remain possible; bit-identical output across processing paths is not guaranteed.

All AI models and their runtimes are included for offline use after installation.

Close OmaRAW, then install or upgrade on x86_64 Arch Linux / Omarchy:

sha256sum --check --ignore-missing SHA256SUMS
sudo pacman -U ./omaraw-0.1.0beta6-1-x86_64.pkg.tar.zst

Continue with installation only when the package checksum reports OK. SHA256SUMS covers the installable package and its complete corresponding source bundle.

OmaRAW 0.1.0 Beta 5

Texture preview and standard noise reduction fixes for this beta.

• Completed edit previews now appear while full-resolution detail is still processing, so older detail tiles no longer hide a Texture change.
• Texture slider changes reuse GPU shaders, and preview intermediates share the configured memory budget more effectively.
• Rapid standard noise reduction mode and slider changes keep the latest requested values. Reset, undo and mode switches discard pending slider samples.
• Updated noise reduction help explains Amount, Keep fine detail and checking results at 100% zoom. Slider strength and AI denoise are unchanged.

Automated local checks cover Texture, noise reduction controls and pixels, preview caching and memory limits, edit ordering, and GPU/display failure recovery. Large-photo Texture processing at 100% can still take several seconds. Intel Arc B580 hands-on validation remains pending; issue #1 stays open.

A strict GPU pixel-equality check after switching standard noise reduction off found a one-level 8-bit difference in 17 pixels of a 512×512 RAW crop. CPU restoration was exact. The noise strength and mode checks passed in both runs.

All AI models and their runtimes are included for offline use after installation.

Close OmaRAW, then install or upgrade on x86_64 Arch Linux / Omarchy:

sha256sum --check --ignore-missing SHA256SUMS
sudo pacman -U ./omaraw-0.1.0beta5-1-x86_64.pkg.tar.zst

Continue with installation only when the package checksum reports OK. SHA256SUMS covers the installable package and its complete corresponding source bundle.

For B580 testing, check Texture at Fit and 100% with repeated drags, Reset and Undo. Check standard noise reduction on a noisy original RAW at 100%: compare Everything at Amount 50 / Keep fine detail 0 with Colour only and Off, then switch modes quickly and move Amount.

OmaRAW 0.1.0 Beta 4

Graphics recovery improvements following reports of GPU memory exhaustion.

• Failed GPU allocations, transfers or dispatches release the affected processing context and fall back to CPU for the session.
• Display-surface failures trigger a controlled restart with CPU processing and software rendering after workers and catalog locks are released.
• A further interface failure exits instead of repeatedly restarting. The recovered window explains the change in its status message.
• Updated built-in help and the user guide.

Local tests cover injected GPU failures, catalog reopening and saved-edit preservation, normal GPU processing, Texture and colour masks. The original Intel Arc B580 workload still needs testing; issue #1 remains open for that verification.

All AI models and their runtimes are included. No additional model downloads are required after installation.

Close OmaRAW, then install or upgrade on x86_64 Arch Linux / Omarchy:

sha256sum --check --ignore-missing SHA256SUMS
sudo pacman -U ./omaraw-0.1.0beta4-1-x86_64.pkg.tar.zst

Continue with installation only when the package checksum reports OK. SHA256SUMS covers the installable package and its complete corresponding source bundle.

OmaRAW 0.1.0 Beta 3

Crop and straightening improvements for the first beta release.

• Auto level is no longer cancelled by preview resizing as its status appears.
• If automatic levelling cannot find reliable edges, Draw level line switches on with instructions. Drag along a horizon or upright edge and release to straighten.
• Draw perspective guides has a clear label, preserves completed guides during preview refreshes and opens when automatic perspective cannot find a correction.
• The Straighten slider has a longer track and a gentler ±10° range. Shift-drag gives finer control; the angle field accepts values up to ±45°.
• Updated the built-in help and user guide.

This package includes all AI models and their runtimes. No additional model downloads are required after installation.

Install or upgrade on x86_64 Arch Linux / Omarchy:

sudo pacman -U ./omaraw-0.1.0beta3-1-x86_64.pkg.tar.zst

SHA256SUMS covers the installable package and its complete corresponding source bundle.

OmaRAW 0.1.0 Beta 2

OmaRAW beta for Omarchy and Arch Linux on x86_64. This package includes all AI models and their runtime libraries: object selection and editable masks, object removal, Bayer and X-Trans RAW denoise, and automatic subject tags. AI processing works offline after installation, without first-use downloads or setup. The private Python runtime is also included.

Download these two assets below into the same folder:

• omaraw-0.1.0beta2-1-x86_64.pkg.tar.zst — the complete installable application.
• SHA256SUMS — checksums for the release assets.

Open a terminal in that folder and run:

sha256sum --check --ignore-missing SHA256SUMS && sudo pacman -U ./omaraw-0.1.0beta2-1-x86_64.pkg.tar.zst

The checksum must report OK before installation. Pacman installs normal Arch system dependencies during installation. No additional models or runtime packages are downloaded by OmaRAW afterwards. The package is currently unsigned.

Close OmaRAW before updating an earlier beta. Open OmaRAW from the application launcher after installation; on Omarchy use Super+Space. No separate AI setup step is needed.

While this repository is private, sign into GitHub before downloading. Browser sign-in does not authenticate the README’s terminal downloader; use the manual installation above.

The package includes the RAW processing engine, film simulations, built-in guide, agent skill and required licence notices. Press F1 for help. Start a new agent session after installation; omaraw skill checks its links and omaraw skill --link repairs them. Photos stay on your computer.

Start with a small collection and keep backups of your photographs and catalog. Camera, RAW, tethering and GPU support depend on hardware and files. CPU processing is included; full-resolution denoising can take substantial time and memory.

The attached omaraw-0.1.0beta2-1.src.tar.gz contains the matching application and engine sources, original model inputs, pinned runtime packages, Python and rawpy sources, and package recipe. GitHub’s automatic source archives contain only the OmaRAW repository.

See the README for first steps, updating, removal and feedback.

OmaRAW 0.1.0 Beta 1

OmaRAW’s first beta for Omarchy and Arch Linux on x86_64. The package includes the RAW processing engine, built-in help, film simulations, required licence notices and the OmaRAW agent skill.

Download these two assets below into the same folder:

• omaraw-0.1.0beta1-1-x86_64.pkg.tar.zst — the installable application.
• SHA256SUMS — checksums for the release assets.

Open a terminal in that folder and run:

sha256sum --check --ignore-missing SHA256SUMS && sudo pacman -U ./omaraw-0.1.0beta1-1-x86_64.pkg.tar.zst

The checksum must report OK before installation. Pacman installs the required dependencies and registers OmaRAW in the application launcher. On Omarchy, press Super+Space and search for OmaRAW. The package is currently unsigned.

While this repository is private, sign into GitHub before downloading. Browser sign-in does not authenticate the README’s terminal downloader; use the manual installation above.

Start with a small collection and keep backups of your photographs and catalog. Press F1 for the guide. Optional AI models download when requested; photo processing stays local. Camera, RAW, tethering and GPU support depend on your hardware and files.

For agents, start a new session after installation. Run omaraw skill to check the packaged skill; omaraw skill --link repairs its user links.

The attached omaraw-0.1.0beta1-1.src.tar.gz contains the complete matching sources, pinned processing-engine archives and package recipe. Use that archive to rebuild; GitHub’s automatically generated source archives contain only the OmaRAW repository.

See the README for first steps, updating, removal and feedback.