Interestingly… one can arrange the working system such that the flight of chromaticities toward blackness and whiteness is inverted; toward blackness will converge to complements, and toward whiteness toward primaries. Leads to some subtly nuanced results. (No attempt at tuning. Some posterization manifesting.)
can you link to the specific post that talks about this particular idea? I might have missed that
what, this?
https://blenderartists.org/t/feedback-development-filmic-baby-step-to-a-v2/1361663/1829
that's just three numbers.
Presumably meant to be angles to rotate the primaries by for an adjusted color space? - You kinda dropped those numbers in there without any explanation or context what so ever. Didn't even share the results you showed here over there, so probably a lot of people in that thread haven't even seen what you did here.
But ok, that makes sense, so you are saying all that needs to be done for this is to try to do really extreme rotations of the primaries before applying the sigmoid. I can see how that would effectively flip the relationship of black and white with respect to color mixing. The high end gets soft-clamped towards the opposite of the primary, and the low end gets soft-clamped towards the primary, so if you sufficiently rotate the primary, those things will switch roles.
And then, after the sigmoid, stuff gets inverted.
It's neat, that that still results in reasonable images.
could be fun to do a movie of how this choice affects the end result, simply sweeping through all choices of angles across like, say, 24 frames or so (so each frame would uniformly rotate the primaries by an additional 15° - more frames would allow for a finer sweep of course)
@kram1032 @nihal_rahman 100%! This is a terrific idea!
The only tricky part would be to illustrate how the chromaticity flights shift when we collapse the angles between primaries, as there are infinite combinations here, and the three primaries’ relationships are complexly intertwined.
Still would be interesting to see an animation done in 3D!
Yeah you could shift three different angles, so exploring the entire parameter space in a smooth way is gonna be tricky. You could definitely find paths thruogh that space inspired by space filling curves or the like though.
Or in a different approach, you rotate them by different frequencies, say, based on sqrt(2), sqrt(3), and sqrt(5), so they drift out of sync and slowly explore the entire space. Would take many more frames tho.
However, even just a lockstep rotation (all primaries by the same amount in the same direction) would be a good starting point as a first visualization. You'd just have to pick an arbitrary initial phase and then go from there.
also, wouldn't there be a state where, like, subtractive and additive "cancel out" or "are in balance"? Like, if you rotate by 180°, you're definitely gonna swap additive and subtractive notions. But what happens around the 90° mark?
@kram1032 @nihal_rahman I don’t think there has been much examination here, at all.
We also need to consider that it’s an additive medium, so most certainly not subtractive. This is on my radar as I’d be interested to see how the trichromatic model fares with actual subtractive-like attenuation of channels.
+
@kram1032 @nihal_rahman Yes.
I’ve been wanting to do a fully subtractive approach in its simplest form for a while. I have a hunch it would be interesting to look at the flight paths of chromaticity sweeps.
Still would like to, but it remains somewhat elusive.
@kram1032 @nihal_rahman The reason that these sorts of completely weird transforms are an important thing in my mind, is that it helps to identify general trends in picture literacy.
For example, I can dump quantal catches to a display medium like sRGB, and for the most part, the picture will be read fine, and it can be read as “a creative choice”.
Some values however, will remain curious or uncanny.
In order to arrive at a general protocol, I strongly believe +
@kram1032 @nihal_rahman - that we ought to be sinking a good deal of effort in finding limits and edges where things are more or less agreed upon as uncanny / weird. Perhaps through the exploration of those edges, trends can be generally found, and possibly predictions made.
Once we wake up from the notion that “ideal as measured” replication is *not* a reasonable ground truth for picture making, many of these easily-discarded-as-creative-choice results take on a +
@kram1032 @nihal_rahman - new and fascinating meaning.
For example, quantal catches dumped as BT.709 / sRGB…
nihal_rahman
troy_s 
kram1032