I found this to be one of those images with WB issues…selecting as shot looks like this for me
And auto WB on the entire image is this …
Yes, the ‘as-shot’ numbers from the metadata worked out to 0.999,1.000,0.999. Here are the various WB tags from the metadata, courtesy exiftool:
[MakerNotes] WB_RGGBLevelsAsShot: 1023 1024 1024 1023
[MakerNotes] WB_RGGBLevelsAuto: 2433 1024 1024 1633
[MakerNotes] WB_RGGBLevelsMeasured: 2433 1024 1024 1633
[MakerNotes] WB_RGGBLevelsDaylight: 2236 1024 1024 1722
[MakerNotes] WB_RGGBLevelsShade: 2583 1024 1024 1479
[MakerNotes] WB_RGGBLevelsCloudy: 2416 1024 1024 1591
[MakerNotes] WB_RGGBLevelsTungsten: 1596 1024 1024 2497
[MakerNotes] WB_RGGBLevelsFluorescent: 1917 1024 1024 2383
[MakerNotes] WB_RGGBLevelsKelvin: 2236 1024 1024 1722
[MakerNotes] WB_RGGBLevelsFlash: 2503 1024 1024 1582
The Auto, Measured, and Daylight preset all yielded acceptable results. I also tried the bog-standard GrayWorld WB mode in rawproc, and I got a decent result, not quite as good as the Makernote Auto/Measured/Daylight…
I was talking about the modern workflow with color calibration so I think the D65 values must be off or the translation of them to as shot but maybe not as you show as shot is also off with legacy. Now that I also looked harder when I was using the area selection to get an average I had been using the spot color match previously which has persistent settings. This was impacting the result from that mode which is as expected so it just seems to be as shot?? I think we need a way to highlight that so you know it is altered from 50% Luma 0%hue 0$ chroma… I have done this more than once now…I guess I need to start doing that in a separate instance…
Not really. It was shot using the Unity White Balance / ETTR method (when I get it right), so the camera jpegs (as shot) come out green. And hence the near-unity factors @ggbutcher shows.
I’m not sure why you would try auto WB here. Black tyres or the Land Rover would be better starting points!
1 Like
No! I didn’t encounter the pink because I had the exposure higher and I wanted the image to show some detail of the front suspension of the car. I think we’re somewhat at crossed purposes!, no worries, and there are only so many hours in the day! In summary, this is where I’m coming from 
-
- I suggested sigmoid might benefit from a white point threshold the user could adjust.
- You said clipping was bad.
- I challenged you to demo that because to me that situation is much the same as sliding the top-right point in the tone curve module to the left a little; and I believe that doesn’t introduce artefacts, at least hardly anything to worry about.
- You then saw the pink highlights and thought that was what I was referring to.
Jethro Tull had a song called “Nothing Is Easy”, they weren’t wrong!
2 Likes
+1 to Jethro Tull
2 Likes
Oh, I feel much better. “As-shot” in modern cameras is usually pretty close, and those numbers shook the foundations of my understanding of the universe… 
1 Like
Nothing Is Easy!
I’m actually going to a Ian Andersson concert next month!
I still do not see the need for a manual (non infinity) definition for a clipping level.
We could lower the contrast of the sigmoid module some if it wasn’t for those burned out pixels but you could also fix that compression using the tone equalizer:
The display transform can’t and shouldn’t do everything for you in the editing. Sometimes I feel like users forget that when they see all the parameters in filmic!
Was this closer to your intention?
Threw me off as well as I tend to use the auto white balance if good enough to save time!
Interesting method for determining clipping in camera!
I sometimes jokingly call myself a ETTL guy as I much rather have noise in my picture than clipped highlights! I think that had been a better alternative to this particular picture, there is zero detectable noise in the image even if I raise the exposure by 6 EV!
5 Likes
Ah it was just a quick check of the response of CC to the image I wasn’t trying to nail WB. I did do an edit on another computer and I did a spot on the license plate of the rover… it wasn’t bad…
Thanks for clarifying your wb setting…that with my stupid sticky color match settings gave me some interesting results for a moment…
Wow!
Fair enough then.
BTW, as an alternative to the Tone-EQ approach, if you really want this, you should be able to put that curve module after sigmoid to add that clipping behavior if you really want it. Everyone tells you that you shouldn’t be putting modules after the display transform, but this is exactly such a use case where you have the freedom to do it and you’re welcome to use that freedom.
While there are cases where I strongly disagree with trying to decouple modules where that decoupling can cause your image to easily fall apart due to too many degrees of freedom, I think this is one case where keeping two modules decoupled is ideal.
1 Like
Yes! What an excellent idea.
Well I have tried filmic v6 with and without chroma preservation modes and sigma with 100% hue preservation on a lot of pictures and as a casual user I have found that you can get a pleasing output from all of them but with different settings of colour balance rgb. It’s difficult to say one is better than the other. Some look better with other pictures than others.
Sigma could be useful to give a hard coded pleasing output from the outset with a simple contrast control so is more appealing for quick use or beginners. Filmic gives more fine controls to play with if the user needs to go into greater depth. I think it would be good to be given the choice.
14 Likes
I was following this thread and had really high hopes for the module. Now that 4.0 will come out soon I thought I’d check out if it made it in.
Very sad to see that it hasn’t. From everything I have seen this looks to be superior to all filmic versions so far (haven’t used v6) and is a lot easier to use. Also a bit strange to see only one dev opposing (the one that wrote the competing solution) having that much weight.
@hanatos I think vkdt has a scene to display transform similar to this (except for the log-logistic curve, only using weibull)?
6 Likes
indeed i took the liberty and copied the function. i had another log/S curve in the past, but it was not smoothly differentiable (which is handy if you want to rasterise constant stroke width signed distance functions, which is how i draw the curve in the dspy output, to derail the conversation). also the CDFs @jandren suggested seemed to make a lot of sense to me in terms of point processes (turning silver halide grains during exposure), and the parameters are indeed very well behaved if you play with them (called them light and contrast).
i’m using aurelien’s hue-constant colour space for hue preservation during the transform.
now i suppose i want another curve module because the CDF here can’t invert negatives (but that’s the only limitation i could spot).
8 Likes
Agreed. Clearly this module is quite popular, so why not include it as soon as possible, and those who don’t want it could simply ignore it when editing.
Because someone has to do the actual work?
1 Like
I thought there was a bit more to it than that. As far as I could see, the feature still had a ‘contraversal’ label so wasn’t being included in the main version. I guess the developers have their reasons for this, and those views have been aired in various places including this forum.
1 Like
there’s no pull request for that based on current code base. So there’s nothing to include 
1 Like
Yes, and I imagine a chunk of hours is needed to update the manual for example. But I did say possible.
The code has been around for a while, seems to work well, is popular, so it seems a shame not to have it fully available.
So @jandren , is the ball in your court?!