@troy_s But sensor size can be baked into the focal length as far as fov is concerned.

@GiantCowFilms Not for barrel distortion or any number of other facets.

@GiantCowFilms Sensor size is critical for many things that are byproducts of non-Platonic reality things. Even lens flares etc.

@troy_s You can flag the comment and ask a moderator to delete it.

@OldMan Blam will give you a very good approximation to the focal length, but keep in mind that any time you are dealing with lenses in the real world you have distortions and the optical center might or might not be in the center of the image.

@OldMan I was getting a focal length of 18mm aprox on the image you linked. But the image also has some barrel distortion that prevents absolute accuracy.

@OldMan Also go figure if the walls are indeed parallel or if the floor is really flat... For an excercise or texture mapping like the one you are doing you should be able to get close enough with blam. But be ready to spend some time calculating the real dimensions from objects based on how they are represented on the lens. In my experience the hardest part is determining the correct camera rotation.

Height you can sometimes figure out by drawing converging lines to the vanishing point. Generally the vanishing point will be at the camera height.

@OldMan A trick I've used is painting the converging lines on top of the image so that I can have a clear view of the vanishing point. Set the image as background. Then I set an empty at the height that I suspect the camera was at and make it the Target of a Track to constraint for the camera. so that the camera is always aiming at that point.

Then by moving the camera and adjusting a piece of geometry that resembles the geometry I'm trying to reconstruct I get decent results...

In any case, it would be easier, in my opinion, @OldMan if you made an image for which you know focal length, camera sensor and height and tried to learn that way "camera mapping"

Start with variables you can control. Then when you manage those move on to unknown variables and higher levels of complexity, not the other way around...

@cegaton Thanks for your advices.Btw ... this time I was referring to another BLAM related question BTW .... yesterday I did precisely what you recommend me (making an image and use that for camera mapping). That is what I asked about on BSE blender.stackexchange.com/questions/66970/… T

Hi troy, centaogn and all other users; i found this chatroom while trying to find answers to the whole color management topic. I am still learning and did not get all things i need, but i'm on my way. First of all, thanks for all the useful hints.

@pixelpoems.de So happy to have another person around. Welcome. If you use the @pixelpoems.de format, the people will get notifications.

@pixelpoems.de Answering your question from the BSE question, I believe that if you are seeking to increase the contrast in your image, or other grading adjustments, my best advice is as follows:

A) Use the Filmic Log Encoding Base view

B) Pick a contrast from Looks that is closest to where you wish to go

C) Use a single CDL node between your output and the viewer / composite out

D) Adjust Power to increase or decrease contrast, in conjunction with a little / slight adjustment to Slope as required. Offset is rarely required unless very aggressive Power is applied.

@pixelpoems.de hi Troy, is this usage of the @ Symbol right?

Thank your for your advice - and yes, what you wrote is exactly what i was looking for: Increasing of contrast in a specific way.

I already used the ASC CDL Node after the Render Layer-Image Output, and after that a RGB Curves node. And this leads my to my first question here:

If it - just in theory - would be possible to add a thousand points to each curve of each C R G B Curve of the RGBCurve node, would i be able to simulate a "Film Emulation"( in example Codak Portra ) look? i learend that a 3D LUT is a set of input-output values that where set into relation to manipulate the contrasts of the image that is manipulated. Am i right with my example to understand what is happening?

@GiantCowFilms @pixelpoems.de Close, in front of the person you are contacting. Use their full handle.

(mine is troy_s)

@pixelpoems.de Possibly, but I haven't looked at it. If it is strictly contrast then yes. If it is saturation / colour adjustments then no.

@tr

@troy_s ok, one more question about this chat: How can i post images here? I'll give you an image of a test render using the Filmic Log Encoding Base, it will render over night. While setting up the scene i got mad at first, but then i figured out that i need to drasticly change the color setting of my shaders. Most colors need to lowered in saturation and value. Is that a good sign?

@pixelpoems.de If you look next to the "send" button you see "upload...", which takes an uploaded image or a URL.

@pixelpoems.de The silly chat needs karma though, so might not be unlocked to you yet?

@troy_s sorry, i do not see such button - i'd send you a screenshot if i could ;-) maybe i need to log out and in again?

@pixelpoems.de Regarding saturation, I always start by trying to explain that the various shaders have real-world analogies... Diffuse shader for example is called "albedo" which is not a colour as it is labelled.

Just upload to pasteall.org and paste the link

Albedo is a measurement of reflectance from 0 to 100%. In terms of reality, very few things get up above 0.8 and below 0.05

Nothing is a black hole, so 0 for example, should be avoided.

It also helps to divide your work up between thinking of your scene as a real-world scene to photograph, and the act of taking the photo for creative work and grading.

@troy_s i already heard about this - avoid full zero as well as full one. I already fixed that in my shaders for the colors, as well as for "functional" colors such as color ramps

So using our above example, you wouldn't put 0 into your scene for albedo, but rather light to ratios then grade those values down to 0 as you would with a photograph.

Typically, most things do not go above 0.8. Pure shiny snow / ice for example.

You begin to get a sense of things that aren't real, as well as orient your entire thought process around scene referred values.

The moment you realize there is a division between scene referred, and display referred, is the moment you begin to become aware of how your pipeline can be adapted to more greatly augment your creative output.

I always try to emphasize that in motion picture work there is a person that earns a living as a grader that you take your work to. Grading is a dedicated phase in a pipeline and embracing it is empowering.

wich already has happend over the last days. Until now i mixed rendering and "artistic" tweaking, which often confused myself. now i understood that it a better way to focus on the scene first instead of jumping into the blender standard "film look" panel all the time in hope to find a solution why the scene is trash.

(You also gain on workflow because you aren't mixing and matching phases that don't belong together, and instead are able to focus your energy like a laser on doing what you are supposed to be doing. Just like the order of events of building a house; you wouldn't be painting a room and reframing at the same time.)

@pixelpoems.de Exactly. You need a guiding baseline to give you an idea of the creative direction you want to go, which hopefully the five contrasts give you some close idea. You can push this further and have a look defined that is closer to your aesthetic creative choice too of course.

@troy_s that sound plausible, and i already felt what that means, at least i think so. i often had teh bad experience that i could not reach the desired result with the scene, so that i had to change my idea of the look. But this is not also a problem of color management, but also of the lighting setup and the dynamic symmetry of the image and so on.

In the end, there is a lot of room for creative growth in process and pipeline. You will also discover confused design patterns exist in Blender from developers that didn't quite understand the difference between the scene referred and the display referred. Lift Gamma Gain for example is display referred, and doesn't work. Most AdobePDF blend modes are display referred only and don't work.

@pixelpoems.de Exactly. This is why I am not very keen on trying to evangelize the concepts.

The whole conceptual framework requires a nuanced eye that comes from experience. I am primarily focused on 2D stills for example, with some very basic 3D, but the conceptual leap was huge

@troy_s i will get deeper into this. Step by step.

I certainly wouldn't have been open to it unless I had rammed many hours into imaging and wondering why things were broken

sorry, i will be of for a while. The render is now finished. i will not do post pro.

now do, not "not do"

Why linearized blends? How does a photograph take a massive dynamic range photo yet my "linearized" data is 0..1?

How do I see the range of my data?

Why do people say whiter than white or "HDR" values? (The answer to this is because they are incorrect.)

The scene referred domain exists entirely outside of the display referred domain, and there is a huge concrete wall between them with haywire and prison guards at the top of it.

Mixing and matching those two countries is problematic at all times.

@troy_s one more question: What is the difference between the combination of "Filmic log encoding base" and Look "Base Contrast" against "Filmic Base Contrast" and Look "none"? To me the results look the same.

@pixelpoems.de Astute. And something I am struggling with.

In the original incarnation, I had it broken down as a clear division between View (a sort of baseline transform) and augmented Looks (creative twists on that view)

If you go over to BlenderArtists.org you will see much angst over the fact that people weren't able to just flip a view on and be "done".

So after discussion with a good number of people, I provided a "bridge" for inexperienced imagers to get up and running with the concepts immediately via views, which were exactly as you noticed, simply "Use and go" versions of the View + Looks

I may change that and in fact remove those views. It is a difficult choice.

The most important thing for imagers to grab onto is the conceptual framework below the creative "Wow this looks better" idea

I worry that by making it "easier" I am also sort of crippling imagers by hiding them from the concepts.

Anyways, it is a decision that may change. Best practice of opinion is to use View + Look.

@pixelpoems.de Does that make sense?

@troy_s , yes, i understand this. If i would be that deep in a topic that seems important to me, i would decide to deliver a good but maybe hard to understand solution instead of a simple solution that declines all chance of improvement of the user. we all know that simple solutions are usually bad ones.

so, here is my first render test using the filmic log encoding base

dropbox.com/s/wo96oc49noflqq0/…

i called i "whitemare" ;-) still noisy due to the los sample amount, but i'm quite satisfied with the result. What do you think? I decided to do a milk and glass thing in black and white to a) test the brightness without focusing to much on colors, and b) because i love b&w white images .

@pixelpoems.de Things improve the more imagers use the concepts.

Couldn't type.

The more folks like you and everyone in here understand, the better it is. More knowledge spreads out and makes it less difficult to explain (also much less resistance)

If you read back through the Blender mailing list, you can find more than a few instances of imagers getting very angry at me for suggesting their knowledge of RGB colour is somewhat broken and that white / black don't exist in scene referred models.

@pixelpoems.de Wow.

@pixelpoems.de Are you happy with how the views and looks are responding?

@troy_s yes, but i found it a bit fizzeling to tweak the sliders, especially on the RDBCurves node i placed at the end of the post pro nodes. Only little tweakings tend to flip everything. the ASC-CDl node sliders are very sensitive too, but the result is worth it. Regarding the image the stains in the bottle could have been better, but they are just quick procedurals.

@pixelpoems.de Curves are broken.

The reason is that when dealing with UI, they need to permit an imager to select a transform.

To think of it another way, we cover 16.5 stops of latitude from minimum to maximum that you see. 10 stops of that is below middle grey.

The curves can be mapped to that upper scene referred value of 16.291, but the curves are operating on the scene referred linear values.

That means if we scale the UI from 0 to 16.291, over 10 stops of critical middle grey to dark values are crammed into about 10% of the left of the curve UI

It makes it impossible to control.

@troy_S ah, that is a good hint.

Anyways, in order to get things like that fixed, it requires more imagers that understand what they are doing, understand what the issues between the scene referred and display referred domain are, and then we can work on getting the nodes code and UI fixed. More people. Larger culture.

Experienced imagers.

Thanks to @cegaton who got the whole ball rolling and started. If it weren't for him, none of this would exist.

@troy_S thank your for giving direct feedback. It was a pleasure to talk to you. For me it is now time to go to bed here in germany.

Night. Also on IRC and email if needed.

Easy to find.

@troy_s i will keep using the new information and will continue; new works will be posted here. bye.