You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
@bowald found that <gltf_colorimage> generated very dark images in 1.38.10 (and 1.38.9). We have investigated and found that there seem to be a regression in 1.38.9 (Possibly d634311).
To reproduce
The requirements for triggering this issue is:
An "inner" node which uses an <image> node and exposes the filename input in its interface
An "outer" node which uses the "inner" node, is multioutput and also exposes the filename input in its interface
In that case shadergen will generate a colorspace transform for both nodes which in our case will apply srgb_texture to linrec_709 twice resulting in a very dark image.
Hi @rasmusbonnedal
Can you confirm if this occurs if you explicitly set the target colorspace in the baking context using context.getOptions().targetColorSpaceOverride = linrec_709
I'm wondering if this is because the getActiveColorSpace is empty.
Description
@bowald found that
<gltf_colorimage>
generated very dark images in 1.38.10 (and 1.38.9). We have investigated and found that there seem to be a regression in 1.38.9 (Possibly d634311).To reproduce
The requirements for triggering this issue is:
<image>
node and exposes the filename input in its interfaceIn that case shadergen will generate a colorspace transform for both nodes which in our case will apply srgb_texture to linrec_709 twice resulting in a very dark image.
repro.zip
Expected result
(From 1.38.8)
Actual result
(From 1.38.10)
The text was updated successfully, but these errors were encountered: