rendering fire with Deep data (no holdout) for comp?

   1622   8   2
User Avatar
Member
28 posts
Joined: Aug. 2006
Offline
Hi all,

I am seeking suggestions as we would like to render fire with Deep output for comp but facing issues in productions.

scenario:
1) Karma as the production renderer. H20.5
2) DEEP output with NO matte holdout setup in 3D/Houdini is preferable. (to exchange artist time of setting up holdout with render time and disk space)

FX dept provides shaded fire as VDB, depending on the shots and type of fire, the fire density can be low or high, which yields following issues:

Issue 1 - lacking alpha or color correction mattes: varying fire density translates to less or more solid alpha. Compositors needs to pull custom mattes to color-correct the fire for desired look. Is there a recommended approach that can be done either in FX or LIGHTING that would provide proper mattes for the fire?

Issue 2 - DeepRecolor and DeepHoldout: when the fire density is low, the DeepRecolor-ed RGB is unusable, which makes writing out Deep for holdout a moot.





So, in productions where Deep output is allowed/preferred, I would like to learn what I am missing in terms of setting up the fire (in FX) and rendering the fire (in LIGHTING) so comp can do accurate holdout and color-correct the fire easier (with mattes provided, NOT pulling luminance matte in comp)?

PS. if the issue has to do with how the pyro shader is setup, please share thoughts like I don't know much about it.
Edited by noicevstime - June 4, 2025 03:56:42

Attachments:
Snipaste_2025-06-04_11-.jpg (90.4 KB)
Snipaste_2025-06-04_14-.jpg (133.9 KB)

User Avatar
Member
287 posts
Joined: Jan. 2015
Offline
Did you try to render with uniform sampling instead of pixel oracle to clean up the grainy deep?
User Avatar
Member
28 posts
Joined: Aug. 2006
Offline
Heileif
uniform sampling instead of pixel oracle
Do you mean switch from Variance to Uniform for Pixel Oracle?

Attachments:
Snipaste_2025-06-05_14-.jpg (29.6 KB)

User Avatar
Member
287 posts
Joined: Jan. 2015
Offline
noicevstime
Heileif
uniform sampling instead of pixel oracle
Do you mean switch from Variance to Uniform for Pixel Oracle?
Image Not Found

Yes.
User Avatar
Member
28 posts
Joined: Aug. 2006
Offline
I did give the Uniform sampling a go. The result is less grainy but still noticeably grainy.
User Avatar
Member
287 posts
Joined: Jan. 2015
Offline
noicevstime
I did give the Uniform sampling a go. The result is less grainy but still noticeably grainy.

Did you try to increase the samples to?
User Avatar
Member
28 posts
Joined: Aug. 2006
Offline
Heileif
noicevstime
I did give the Uniform sampling a go. The result is less grainy but still noticeably grainy.

Did you try to increase the samples to?
The primary samples or there is way to set the sampling for deep?
I have lowered the DCM compression to 0 but it's not helping.

Attachments:
Snipaste_2025-06-07_15-.jpg (21.5 KB)

User Avatar
Member
287 posts
Joined: Jan. 2015
Offline
Primary samples.
User Avatar
Member
28 posts
Joined: Aug. 2006
Offline
I have comprared increasing primary samples and it is not revolving the issue of grainy DeepRecolor'd RGB.
  • Quick Links