« first day (390 days earlier)      last day (578 days later) » 

GiantCowFilms
GiantCowFilms
00:22
@troy_s But what if your lowest value is zero?
troy_s
troy_s
@GiantCowFilms When is your lowest value zero in a raytracing engine?
A good example. Close your eyes. What is the dynamic range of what you are seeing?
Maybe a little translucent light through your eyelid... All the way down to... Uh... Zero.
Zero is infinite.
As would be infinity up.
Those aren't values
In a photo it can be trickier due to noise floor, but the essential core concept is the same.
David
David
00:51
@troy_s encase you have not seen it yet: blenderguru.com/tutorials/secret-ingredient-photorealism not a bad article.
troy_s
troy_s
01:03
@David My Twitter and email uh... Filled up.
@David Not sure that sort of growth is going to be super healthy. Happy that the key points are in view on the subject of nodes and things that need fixing. Lots of confused folks too.
@David Likely means more effort on community sites to sort out the confusion.
10 Replies
10 Replies
@GhostEcho Ask in the renderfarm

 The Renderfarm

A place to talk while we wait for our renders to finish. Site ...
David
David
01:41
@troy_s lol. I'm always the last to share new news.
troy_s
troy_s
@David It is apparently going a bit bonkers.
cegaton
cegaton
02:38
@troy_s the biggest issue is that now you have to deal with "alternative" facts..
2
troy_s
troy_s
@cegaton Holy cow yes.
@cegaton Some folks actually are doing a great job against the alternative facts.
@cegaton Only frustrating thing is how confused / confusing colour is when you have a Terra firma foundation
I get that folks have trouble with it, which makes me angry when you see so much rubbish out there.
Here is a classic case in point. It is a comment that is really "Ho Ho ho... So happy everyone else is finally getting around to this thing that I have known about for, well... Good on you guys."
"The real photo though :P The only thing that looks real is the green chair and the wrinkles are what gives it away.

Seriously though. It's called Color Space and not Color Transform (Not sure why Blender would put a custom name to it instead of the most known). And it is interesting that you went for a Log profile instead of Linear. Maybe because nowadays descend cameras have also a Log profile and it's easier to match with it instead of rendering in Linear.

Cinema 4D went to using by default Linear Color Space since about 4 versions ago.
Honestly, these folks drive me nuts.
Middle after confusing muddle.
(That above is a YouTube copy paste)
Minus the last two comments that are mine.
Specifically, Price originally used the term "Colour Space" and I corrected it to make it clear. Here this chump decides that he is going to "Show us the light".
Such hubris.
Anyways... Hopefully there are enough well equipped Special Colour Forces peeps out there to win the skirmishes and conquer the misinformation.
10 Replies
10 Replies
Holy ranting batman!
I'm glad andrew price will listen to people with more experience and better understanding of things. He completely redid one of his PBR material videos when someone pointed out that some of his analysis was wrong. Props to him.
troy_s
troy_s
02:57
@10Replies It is pretty crazy. The goal has always been to try and get everyone on the same page so that folks can see why things need to be changed. The issue is that in this case, it is a massive shift of mental model.
@10Replies It is hard enough for someone to sit one and one and try to iron out the hiccups. The smog of "Please please please listen to me" is frustrating in that it creates more confusion.
Hell hard enough to check a presentation for errors. They slip in. See his "Multiply" comment, which is one of the few blend modes that works.
 
2 hours later…
GiantCowFilms
GiantCowFilms
04:46
@troy_s Fair enough. What should one use for the minimum then? How did you determine those numbers. I want to get my brain around this.
troy_s
troy_s
05:23
@GiantCowFilms It is quite easy to figure out dynamic range given a bit depth. In the case of display referred imaging, maximum is 1.0 and minimum is 0.0
@GiantCowFilms So at 8 bit, we can simply assign the minimum value as the smallest representable fraction (1/255 for example)
troy_s
troy_s
05:38
@GiantCowFilms log(high)-log(low)/log(2)
@GiantCowFilms So, (log(1.0)-log(1/255)) / log(2)) should solve the question of sRGB or any other ranged colour space when linearized.
@GiantCowFilms Make sense?
sRGB is 7.9943534368588
If we encode the full range of 8 bit data, we would end up with a perfect 8 stops.
1/255 assumes we encode 0.0 as 0.0. If we used a proper lookup to encode that 0.0 value to the smallest encoded value, the lowest value would be 1/256.
GiantCowFilms
GiantCowFilms
I see
that makes sense
 
11 hours later…
Mareck
Mareck
16:41
So finally here we are, Filmic under the spots lights. Hope he will correct his mistakes, I was WTF when I saw multiply isn't working.
And I think he didn't understood transfer curve. He spoke it in the way it's to decode binaries data from the camera to display images... the good thing is that now people know that there are broken tools.
@troy_s hope you wasn't attached to your mailbox ^^
troy_s
troy_s
@Mareck It is so hard to get every single bit right. I corrected a few of those, it is going to be expected a bug or two would creep in. He was pretty keen on getting the base concept out, which traded off some of the technical nuance.
Mareck
Mareck
And it isn't an easy topic
Even me (not that I'm mastering the topic) I've already said bullshit, at least not right assumptions.
troy_s
troy_s
The surface area, when you refactor the colour understanding back into the mega-complexity of shaders and rendering and on and on is huge.
Hence it takes a village.
Some of the funniest stuff is in the comments.
Mareck
Mareck
I'll go read them
Mareck
Mareck
17:01
Very heterogeneous, some spotted that now they don't need tricks anymore while other haven't catched the point.
But I think we'll see increasing of quality renders in next futur.
My colleague has done some nice ones, let you know if he post them.
 
7 hours later…
troy_s
troy_s
23:32
@GiantCowFilms The math on sRGB should end up quantizing that lowest value of 1/255 to 0.000306blah in scene referred, so that amounts to around 11 stops or so. Most of it is biased towards the extremel low values.
10 Replies
10 Replies
We just need one more person, If you have any points in the blender tag on SO it would be nice if you could contribute stackoverflow.com/documentation/blender/commit

« first day (390 days earlier)      last day (578 days later) »