[ENG] Filmic and Sigmoid (part 1)

[@bernym4047]

Glad to see you making videos again. A very well explained concept. Thank you.

[@adabbleinphotography8721]

Thanks Berny, glad to hear from you again!

[@Eigil_Skovgaard]

Almost missed this. Looking forward to 2.

[@adabbleinphotography8721]

Thanks Eigil, don't hesitate to subscribe!

[@michalkvasnicka3290]

Awesome. Many thanks. Looking forward to the next part.

[@adabbleinphotography8721]

Thank you, lots more to come!

[@emmypuss4533]

Good explanation - thanks

[@adabbleinphotography8721]

Thank you.

[@tonybarrett7557]

Excellent

[@adabbleinphotography8721]

Thanks

[@jscfos]

C'est un plaisir de vous retrouver pour vos vidéos sur Darktable. Votre méthode pédagogique est vraiment excellente. 👏 Bravo.

[@adabbleinphotography8721]

Merci beaucoup.

[@olivierparlephoto]

In Affinity Photo, in the Photo Persona, you have set a sRGB/8 colour space, and you are right, Exposure will move the photo in the display scale you have shown. Develop Persona however is using the unbounded 32 bits linear scale you draw (the scene). If you set the colour space of the Photo persona to Pro RGB / 32 linear, Exposure will react as in the Develop Persona. I guess this is the same for Camera RAW's Photoshop. There is a nice explanation by James Ritson (Affinity Photo) about it. Also, you left Exposure using luminosity and colours. I usually change the fusion mode of the Exposure layer to Luminance, that way colours are not affected. This is also what Filmic's developer suggest to do when using Filmic module. In short decorrelate luminosity from colours. This is a complex task to explain ! so congrats you try :)

[@adabbleinphotography8721]

Thanks Olivier, I am using voluntarily Affinity in 8bit display referred to show what is it. I am aware raw apps (like the develop persona) are different beasts. That is the point of part 2. I did not want to separate colour from exposure because the whole point is to show that colour conservation is an important part of developing a photo. Increasing exposure in a luminosity (or such) fusion mode has no relation to physical reality, which is the whole point of a scene referred workflow. I'm interested to have a link to Aurelien Pierre (the developper of filmic in darktable, not the creator of the theory) stating that we should separate luminosity from chroma. It seems to be the exact opposite of the principal of the module! Thanks for taking the time to comment!

[@olivierparlephoto]

@@adabbleinphotography8721 Oh, ok :) I have certainly missed the moment you said you did that voluntarily. Sorry ! Can't wait for part 2 then :)

[@adabbleinphotography8721]

No problem, I probably lost some clarity when I cut the subject into separate parts. I actually didn’t test the develop persona for colour/data conservation. The idea is to increase exposure by let’s say 3 stops, the oranges should stay in the same hue. Then with another slider/module reduce by 3 stops and see if we come back to the starting point.

[@rosleinrot5322]

This topic is getting me! In dt I work with the scene-referred mode, because some tutorials did recommend so. But don't know exactly why. You're always doing very good explanations, can't wait to see the next video! PS.: Love to learn some extra skills on Excel ❤

[@adabbleinphotography8721]

Thanks, we have the same questions! My skills on excel are not great, I use basic functions with my students. Thanks to the person who wrote the macro I used!

[@andrewlarkin200]

My question is why adding/subtracting a constant value to each channel in Excel is expected to *not* change hue (irregardless of channel clipping). I would have thought a proportional adjustment would be the accurate model - add or subtract a percentage for each step. Channel clipping is obviously an issue but I feel your Excel explanation is flawed.

[@fdyocum]

The best explanation I have seen of the limitations of a display referred workflow. Looking forward to the next video!

[@andrewlarkin200]

​@@fdyocumare you replying to my question or commenting on the video itself?

[@adabbleinphotography8721]

A better way would be to transform RGB to HSL and play on the luminance, but I think that you understand that this wasn’t the point of the segment. My very basic method was enough to predict the change in hues and to show what I wanted.

[@andrewlarkin200]

@@adabbleinphotography8721 my point was that a change in exposure would affect the channels proportionally up to the point of clipping. Without this proportional change, there is an inevitable change in hue. Adding 10 to each channel is only valid for greys. Working in a non-linear colour space makes this harder as simply applying a percentage change assumes linearity and again will change hue. I am not saying the hue-shift effect you are discussing is wrong - just that this part of the explanation of what is causing it is flawed. Channel clipping will definitely cause a hue shift as the channels can no longer be adjusted proportionally. This is the reason for using scene-referred workflow. But the model you have explained via Excel changes hue before clipping even occurs.

[@fdyocum]

@@andrewlarkin200 Apologies, to the video.