Hello, thank you for the very interesting project ! ( just discovered it )
In the past i experimented to use Darktable for a scene referred workflow with an OpenEXR sequence rendered in Blender.
After color grading in Darktable a single OpenEXR image, it was possible to copy and paste the “Style” to a sequence of images.
But of course Darktable was not designed to manage video and the process was slow and not very productive ( no video playback for example ).
So would it be possible to work with OpenEXR sequence in vkdk ? ( and maybe export in OpenEXR, TIFF or FFMPEG for video compression)
For what i understand vkdk started to support video too and it would be great to work with OpenEXR sequences.
For now i have difficulties to find a productive free and open source software in GNU / Linux to work with color grading in a scene referred workflow.
For example not all Blender compositor nodes are compatible with scene referred.
As soon as possible i try to experiment with vkdk with different distros if compatible with my GPU ( desktop with RX 580 8 GB vram and notebook with A730m 12 gb vram )
right, vkdt has some basic video support. it loads the .mov my cameras produce, as well as .mlv, magic lantern raw video. audio support is terrible to say the least. as of yesterday or so you can select the different prores presets for 10-bit video output from the gui.
vkdt will support the workflow you describe (edit one and copy/paste history), or you can give it a sequence of images to work on, say IMG_%04d.CR2 and it will interpret these as a timelapse sequnce and allow you to play back in darkroom mode and put keyframes on your module settings.
(edit: see the i-raw docs here which apply to other input modules too, and maybe this video for a screen capture how to produce timelapses when using vkdt from obs or git)
both your gpus should work more or less well. the only thing being colour pickers and stuff like that depend on atomic float addition on images, which seems to be only supported on nvidia (i have workaround code in place, but it’ll be a bit slower).
i haven’t yet implemented any openexr support. it’s a biiiiig collection of libraries as a dependency and requires c++. i’ll probably be open for read/write support via tinyexr. this is probably a sunday afternoon job.
sure. test data is always good. especially since exr can store pretty much anything it would be good to see what exact encoding (single layer multi layer, float half uint, scanline or tiled etc) you’re using.
For i now i can share a super basic sequence test ( single layer float Half ZIP lossless ) rendered in Blender Cycles .
When possible i try to share a more useful “production like” multi layer sequence.
