vkdt nightly build 20241210

this is an automated nightly build of vkdt.

it may contain pre-release quality features, use with care :)

appimage

the appimage compiles in a lot of features
(mlv, v4l, asound, i-vid, o-vid, exiv2, freetype)
but not all
(no quake).
you also currently need the exiftool binary installed on your
host system to enable all features (metadata handling).

appimage with rawler

use this for CR3 raws (newer canon cameras) and for older ubuntu
that don't have a recent libstdc++, uses the rust-based backend
instead of the c++-based rawspeed which does not have CR3 support.
also it builds a custom glfw with pentablet/wacom support

appimage with rawspeed

these are currently temporarily disabled due to some dynaming linking error
when running the appimage.

appimage with pentablet

links against a custom version of glfw, enabling drawing of masks via pentablet/wacom device.
only available in xorg version currently.

windows with rawler

the windows version is pre-alpha and has known issues, see #103 .
it compiles less features
(freetype mlv, i-vid, o-vid)
and leaves out many
(no v4l, no sound, no quake, no pentablet support).
you can help by reporting more issues (or maybe even fixing, if you can ;) ).

0.9.1

0.9.1

this is a minor point release over 0.9.0 including a few bugfixes, especially for scaled windows on wayland.

new features are only on the master branch.

0.9.0

this is a feature release of vkdt with mainly architectural changes in the background as well as ui polishing as focus.

note that due to several hard drive failures currently vkdt.org is down and the documentation can be accessed online here instead for time being.

user interface:

• gui now based on nuklear (instead of dear imgui)
• a lot of nuklear/gui streamlining fixes, bringing back:
  • some keyboard navigation
  • some gampad navigation
  • dual monitor support in graph view
  • style.txt contains ui colours for customisation, as always, the one in ~/.config/vkdt/ takes precedence
• zoomable graph editor (can finally work with large graphs, yay!)
• some new hotkeys added ("upvote and advance" etc)
• re-did the collection interface in lighttable, streamlined support for per-day rating workflow
• added lighttable header describing the visible collection
• recently used collections now also store filter attributes
• asynchronous image rendering and gui display, makes ui feel more responsive on slow machines

processing modules:

• colour module now reads and applies the camera as shot neutral white balance by default
• denoise module with detail protection slider, especially useful on low-res telephone images
• MotionCam .mcraw video files will not by default apply the embedded gainmap since it is
  unclear which raw files have it pre-applied (so switch on in the denoise
  module if your files need it)
• o-web: new transitional module that delegates to o-jpg and o-vid, so you can export
  images and videos of a certain session in one pass.
• usm: unsharp mask, very simple and efficient basic sharpening

tooling:

• vkdt mkssf now considers the HueSatMap in the dcp profile it converts
• support focal length ${flen} in string expansion when loading luts, enables colour profile per lens in smartphones
• vkdt scanin to convert argyll .cht charts to colour picking/colour correction presets
• vkdt gallery now supports jpg and mp4 videos in website creation

platform compatibility:

• vkdt now builds on macos, though there are issues with newer machines
  due to some spir-v/metal translation layer.

0.8.0

vkdt now has a well working nightly package generator running here on github, hence i find releasing snapshots less important. this is the last release which uses dear imgui as widget library, so i want to freeze the state of packages with this release in case the switch has unexpected bumps. the new version will use nuklear as widget library, and is running pretty well. there are two main features that will go missing (at first, maybe):

• gamepad input support
• dual monitor mode

so if you depend on one of these, please speak up and support in the new versions might come back. this release still has them.

main features

• neural network infrastructure. better loss node with
  logarithmic plotting. kpn: kernel predicting neural network for
  denoising, training included as kpn-t. based on tiny mlp using
  tensor cores/registers/shm.
• dngop gainmap support (the denoise module applies it if it's in the metadata)
• colour module with emergency highlight clipping option
• o-jpg can now write exif metadata if requested

new modules

• mask create data-dependent masks from luminance, hue, or saturation of an image
• mv2rot compute rotation/translation from per pixel motion vectors
• o-exr write openexr images with chromaticities attribute
• o-csv dump image data to text file
• i-mcraw input motioncam raw video, with gpu-based decoder suitable for real time
• grain add parametric film grain with standard deviation and correlation
• format generic pixel format converter, necessary sometimes

legacy modules

• cnn: renamed to resnet (training in gmic may be incompatible now)
• menon: superseded by rcd mode in demosaic module
• o-ffmpeg: removed, superseded by much faster o-vid (with 10 bit/prores support)

platform compatibility

• move platform dependent code to fs.h and implement some windows stuff
• windows package build in ci

other

• rawler/rust-c now the default raw loading backend
• better audio/video support
• reworked colour management, buffers can now annotate their primaries and trc
  front to end, with support for the most common set (DCI, HLG, ..)
• new presets

plus of course bugfixes and code cleanups, gui streamlining things.

Thanks to these folks for contributing!

• @paolodepetrillo for video support improvements and gainmap support
• @blitzgneisserin for icc streamlining, exif work and new presets
• @dydyamotya for file system support improvements
• as well as the motioncam folks for releasing their raw decoder and lots of valuable input for vkdt!

0.7.0

lots of small fixes for complicated/non-directed graphs, ui streamlining and polishing,
doc and website updates.

lighttable:

• add button to create video from timelapse sequence of raw files
• output modules all show their individual parameters now
• new automatic labels: video and bracket

darkroom:

• new dopesheet widget to manage keyframes
• animation controls
• introducing keyaccels: quickly bind small scripts to hotkeys (for instance
  expose up by half a stop, rotate ccw by one degree)

new modules:

• autoexp new module for autoexposure, most useful in video sequences
• i-exr initial openexr file support for reading
• new module hotpx to remove stuck pixels in bayer images

modules:

• hilite smooth out c/a artifacts at boundaries of overexposed regions
• o-ffmpeg support prores PQ/smpte 2084 10-bit output
• i-vid support 420, 422, and 8-, 10-, 12-, 16-bit input
• quake with more parameters and amd raytracing support
• i-pfm supports animation frames and noise profile arguments
• pick colours with freeze support for copy/pasting of settings
• grade with colour wheels input
• blend with option to invert mask
• demosaic now with RCD support. it's slower than the gaussian splats. some
  numbers for 16MP: gaussian splatting: 1.5ms, new RCD: 2.5ms, darktable's
  opencl implementation on the same image: 21ms.

build:

• i-raw update rawspeed to upstream + cr3 (but will not do it again)
• support rawler/rust backend to load raw images as replacement for rawspeed/exiv2
• initial .AppImage package provided

sha256sums:
4a4fcaf845af241930c03e51d601fd6501deb688eb2534c9aca5bd0efafe9470 vkdt-0.7.0.tar.xz
99fc70a4300c0e470dc86af37163b4066cef0a7a1e9f5a3b9b903a18382379b9 vkdt-0.7.0-x86_64.AppImage

0.6.0

this is a feature release of vkdt. the changelog over the 0.5.x series includes:

• github actions: test build on linux, codeql to find potential security issues
• code cleanup/security fixes
• perf overlay to show run-times of animations in the gui
• gui fixes and finetune things
• clean shutdown (always write config file!)
• new graph editor view / removed the old graph edit tab in darkroom mode
• initial focusstack blend mode
• new autocrop button to detect and crop away black borders
• preset for run-time gui noise profiling
• frame-limiter for power saving
• module graph editor coordinates stored in .cfg (see d703743)
• libvkdt build option for 3rd party software, option to build cli only
• dual screen support (display full image on second monitor)
• zoomable mini-displays in right panel
• key accels for rating/lables now configurable
• custom font support for non ascii characters
• quake with path guiding for direct illumination

also see the repository for camera input device profiles and noise profiles:
https://github.com/hanatos/vkdt-camconst

as well as the module docs:
https://vkdt.org/src/pipe/modules/readme.html

if you have questions or issues you can try to get in touch via the forum or in irc #vkdt on oftc.net.

$ sha256sum vkdt-0.6.0.tar.xz
02ed12fbd63e1f416ea0765a7273786b3405401d2d6ab4f66c736c2f1ba37cbc  vkdt-0.6.0.tar.xz

the tarball is signed by key 25C8A341700B4E1D on hkps://keys.openpgp.org and you should be able to check it via

gpg --recv-keys 25C8A341700B4E1D
gpg --verify vkdt-0.6.0.tar.xz.sig

enjoy!

0.5.4

this is a bugfix point release: compared to 0.5.3 this:

• fixes an assertion triggered for mismatching PushID/PopID in the files view
• enables compilation on ARM
• fixes image barriers for the mesa/amdgpu vulkan driver

0.5.3

this is a bugfix pointrelease and fixes a crash when creating a new tag.

as always, to report issues or ask questions, try the forum or irc #vkt on oftc.net.

0.5.2

this is a bugfix point release: compared to 0.5.1 this:

• fixes copy/paste history hotkeys in lighttable mode
• fixes feedback connector creation in node editor (for animations, hold shift to connect output from last frame to input of modules)
• fix spurious crash in node editor when disconnecting nodes
• fix atomic support detection for amd devices
• fix backspace hotkey in files view to go up one directory
• allow to enter darkroom mode even for broken cfg files
  (so the user can go and fix it in the gui)

to report bugs or discuss, try this forum
or irc, #vkdt on oftc.net

sha256sum vkdt-0.5.2.tar.xz
d5b2a765a0266175323f5bb4fb622f3f27db7545c836c451e26d1046af0af73f  vkdt-0.5.2.tar.xz

0.5.1

this is a bugfix point release: compared to 0.5.0 this:

• fixes noise profile generation for high-res images (quick fix)
• fixes thumbnail generation for directories that have not been loaded before

also new:
the camera constant/noise profile/input colour profile lut repository:
https://github.com/hanatos/vkdt-camconst

and the irc channel: #vkdt on oftc.net

sha256sum vkdt-0.5.1.tar.xz
20c092e9b26a3b33de00567242d8f3775ad0e6e8087afaa81545ba634c353f3b  vkdt-0.5.1.tar.xz