Yes this is the danger 
6 Likes
I think this is one of its major strengths. You can do a lot just in this one module and therefore skip others.
This is exactly what I did when Sigmoid came out, and I’m already getting the urge with AgX.
8 Likes
The test was successful - but I cannot make a screenshot that shows the tooltip for some reason.
So I think you can merge.
I don’t see an open PR.
I’ll only have time to merge this tomorrow, so no hurry.
ah - ok, I forgot that, thanks
Hi,
A new build with the changes from @betazoid (thanks!) and updated from the master branch. No changes from me – I’ll work on the promised stuff on the weekend.
@Dave22152 , @MStraeten , please update your builds if you can find the time.
https://tech.kovacs-telekes.org/dt-agx/Darktable-5.1.0%2B685~g17ddf0ca22-x86_64.AppImage
3 Likes
Looking forward to some video guide to this. There seems to be no way for me to unclip highlights and get a nice roll-off. White relative exposure doesn’t take me nearly as far as I’d need to go.
Hi, I’m out of town this weekend. @priort could you compile the Windows build?
Thanks!
Sure I will do it now…
EDIT
3 Likes
I mean you’ll never be able to unclip what is clipped… but you should post a problematic image in this thread. The highlight rolloff is like the best thing about this module.
A new build with a few minor enhancements:
- the pivot is now marked on the plot; you cannot drag it, though
- there is a picker next to the pivot X shift; selecting it allows you to pick an area, whose average will be used to set the pivot shift. The code tries to adjust the pivot Y (output) to maintain the brightness of the selected area, but no guarantees. I’ve seen some glitches (completely wrong colours), but have not had time to track them down; if you could find a repeatable way to reproduce them, please let me know.
- the black auto picker will use the darkest of the selected RGB components as the black point. I found this useful, e.g. when I had a red rose on the image, with the blue component close to zero. It makes sense not to push the darkest red to 0, but rather the blue.
- I’ve tried to add gamut protection against negative values due to oversaturation (there is no protection yet against values > 1, please reduce brightness or use other tools in such cases). With very strong LED lights, this can cause the parts of the image to go black (all input coordinates negative, i.e. out of the Rec2020 gamut). This is an experimental feature, and is based on my understanding of @flannelhead’s ‘low end compensation’ (could be wrong, please check): both at the module’s input, and after the AgX code, I try to take care of negative values. If any value is negative:
- the luminance of the pixel is calulated
- the most negative pixel is subtracted from all components to push them to 0 and above
- the luminance of the adjusted pixel is calculated, and then scaled to match the original
- at the module’s input, this is done in the working space
- after AgX, however, the pixel is first converted into the output space, and the check and correction are done there, then it is brought back to the working space.
- you’ll find this in the advanced section:
- example:
The Linux build:
https://tech.kovacs-telekes.org/dt-agx/Darktable-5.1.0%2B723~g6376aa7b05-x86_64.AppImage
@priort , @MStraeten , please update your builds if you have time to do so. Thanks in advance!
8 Likes
I’ll be home in about 21/2 hours…I spin one up then…thanks for the update…
4 Likes
the automatic adaption of pivot y can mess up the edit, if a custom area should be selected for Pivot x.
scenario: you want to autoadjust for a skin tone in a picture with quite dark background.
in dependence of former set selection area pivot x and pivot y are improperly changed affecting the brightness of the whole image. But that will be the reference for adjusting pivot y after resetting the picker area …
Yes, thanks. I’ll see if I can figure something out. Maybe it was a bad idea.
I haven’t really followed the thread too closely (it’s honestly really long and a bit technical for me), so I wanted to ask… is there any chance this would end up appearing in 5.2.0, or is 5.4.0 more realistic?
I’d excited to try it out but don’t know what state it is in right now.
What os?? I’m building the windows build it will be up in a few min??
EDIT:
Win version…
2 Likes
idea is good - maybe better the initial pivot y value when activating the picker needs to be taken into account, not the last calculated
or maybe a toggle to enable/disable that e.g.
Definitely not 5.2, sorry.
1 Like
that discussion is fruitless. it will be in a release if it’s ready at feature freeze for that release. It’s not yet a pull request in the main darktable repository …
1 Like
No problem, I’ll wait