[Bf-cycles] Some suggestions for nodes' interface revision in PBR times

[Troy Sobotka]

On Sun, Dec 10, 2017 at 12:23 PM Matthew Heimlich <matt.heimlich at gmail.com>
wrote:
> It definitely sounds like output formats are being confused with texture
authoring formats here.

Not at all.

You are talking about linearization of a transfer function here. While that
is part of the issue, that isn't the whole thing.

When authoring for wider gamut references, you have to take into account
the reference primaries of the source. For example, every camera out there
today can capture well larger than the REC.709 gamut. It is quite trivial
to generate a wider gamut texture for example.

If you are mastering to a wider reference space, the material is
constrained by the reference primaries it is in. It is rather like making a
high quality digital recording of a cassette tape (for those of us out
there that know what that is).

TL;DR: The assets need to be generated in at least the reference space or
larger.

> Mari, Substance, and Quixel all suggest (and in some cases even force) a
linear -> sRGB workflow.

This would be ridiculous. It also ignores the fact that you can't get to a
scene referred linear via sRGB, so forcing this via some baked in hard
coded math would be absolute, abject failure.

Further, there's a good deal of confusion around what the sRGB OETF is and
how it relates to colour management. Not going to get into that, but I will
say that as more folks understand the impact on the work, the more that
move over to re-generating assets required for 2017 era input and output.
Remember Tears of Steel? That work began in 2011 and in fact rammed face
first into these issues.

With respect,
TJS
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.blender.org/pipermail/bf-cycles/attachments/20171210/568164ac/attachment.html>