Spektrafilm darktable module implementation discussion

Doug-Phoenix · October 3, 2026, 5:15pm

I gave this a go late yesterday - my first time running a nightly. Exports here.

I must say, the module design and layout is excellent! Over the past few months I’ve been experimenting with other spektrafilm modules (vkdt, spektrafilm OFX, and the original/experimental versions of the original). That exposure was helpful. I also have several presets from @bazdesh for the experimental branch, and I have been learning from those.

I am very impressed with this module. Thank you!

ho4 · October 3, 2026, 7:02pm

Now with the black level correction as well as the pre&post compression boosts this module really feels like a complete package. Especially the way the preflash interacts with the boosts to create these soft transitions into the source of light, very nice. I seldom use any other modules for most of my edits now, and if I do, it’s mostly for white balance reasons or selective colour adjustments.

Thank you for the work on this implementation. I think the overall experience using it turned out quite nice

Roberto_Abella · October 5, 2026, 5:22am

For some reason, the colors are not the same when exported. I’m using 5.7.0+1162~gd53157453e

piratenpanda · October 5, 2026, 5:23am

is it spektrafilm specific or a general darktable issue? Could be related to Color Calibration broken on master · Issue #22496 · darktable-org/darktable · GitHub

Is you image viewer color managed? and is darktable as well? So many possibilities

Can’t reproduce

Roberto_Abella · October 5, 2026, 3:31pm

Couldn’t say… if I use Agx the colors are ok, and the same with Sigmoid. I’m on Manjaro linux, and I use gwenview and Viewnior. Neither of them are configured to manage colors (well… not sure about Viewnior but I think is not). May be it’s related to that Color Calibration issue… first time I noticed was last night

mino · October 5, 2026, 3:51pm

Could you upload RAW+xmp so others can try to reproduce?

ho4 · October 5, 2026, 7:01pm

Not quite sure why but seems rather odd:

set post-compression scale to 0.90
use autopicker on pre-compression boost
it sets the value to max (4.00)

Yet:

set post-compression scale to 0.95
use autopicker on pre-compression boost
it sets the value normally (1.45 in my case)

Maybe intended behaviour? Idk about the math side of this

piratenpanda · October 5, 2026, 7:09pm

Will take a look. I have a round of Bugfixes waiting anyway

bazdesh · October 5, 2026, 7:45pm

Could this be on darktable? The autopickers in negadoctor sometimes do absolutely crazy stuff and come up with totally random values.

piratenpanda · October 5, 2026, 8:21pm

I found something. Maybe it fixes this problem in spektrafilm. Will submit a PR with bugfixes tomorrow hopefully.

Roberto_Abella · October 5, 2026, 9:07pm

Sure!
_MG_3886.cr2 (21.1 MB)
_MG_3886.cr2.xmp (10.9 KB)

piratenpanda · October 6, 2026, 4:24am

works fine for me

Roberto_Abella · October 6, 2026, 4:31am

Hmmm… so there must be something wrong with my dt or my OS or something else… I’ll investigate. Thanks a lot!

piratenpanda · October 6, 2026, 4:32am

or the test is not valid, maybe I’m missing something. I’m on wayland, might make a difference

Roberto_Abella · October 6, 2026, 4:34am

Well… I’m on X11 with xfce… but I think there must be something else…

piratenpanda · October 6, 2026, 5:28am

a fix should be in Spektrafilm bugfixes by piratenpanda · Pull Request #22524 · darktable-org/darktable · GitHub

github.com/darktable-org/darktable

spektrafilm: measure the boost picker without the post-compression scale

Commit by

piratenpanda - Spektrafilm bugfixes

master ← piratenpanda:spektrafilm_bugfixes

sf_sim_probe_lightness() switches the compressor off to measure what the pre-com…pression boost feeds it, but left out_scale in place, although that scale is applied after the compressor. Where the scanner black/white point is fitted, the delivered Y is clipped at 1, so the probe's lightness cannot exceed out_scale^(1/3): below 0.97^3 = 0.913 the picker's 0.97 target is out of reach at any boost, and the bisection ran to the slider maximum. A scale of 0.95 picked a normal boost and 0.90 picked 4.00. Hold out_scale at 1 in the boost probe, so the boost no longer depends on the scale set after it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PUcQsmsKTfK2SRKVobpXPs

ho4 · October 6, 2026, 7:49pm

Another bug:

Enable spektrafilm or reset the module to its default state
Change film stock to Tri-X
Change film stock to Double-X
In the “print” tab, the “development time” slider is not available

piratenpanda · October 6, 2026, 8:05pm

Are you running the 0.3.4 pack by chance? Or is this master without changes?

piratenpanda · October 7, 2026, 6:41am

should be fixed with Spektrafilm bugfixes by piratenpanda · Pull Request #22524 · darktable-org/darktable · GitHub

bazdesh · October 7, 2026, 7:40am

I am not 100% sure if this is a bug but - the threshold section in halation is active even when the halation is disabled. Also, the halation strenght and halation size basically do nothing when the threshold is pushing the image hard - it sort of feels like they are in the wrong order in the pipeline, that its halation → whatever goes on in threshold instead of the threshold boosting → halation?

Forgot to add, I am on todays nightly and the 0.3.4 data pack.