PDA

View Full Version : Radiosity issues-good radiosity in VPR, poor results when final rendering.



prometheus
09-01-2014, 01:25 PM
Well..the radiosity gave up on me in this scene..looks as I expected radiosity to effect the objects in the VPR, but final rendering it in perspective cam mode, it doesnīt kick in properly.
using lightwave 11.6.2 32 bit windows 7

there are two lights... two dome lights, one with a small angle to have a harder sun with slight soft shadow edges, one dome light at 90 degree angle for overcaster sky light, and montecarlo radiosity, check the small image and how little of the radiosity is affecting shrox Dante space ship, while in VPR in the big screenshot...that is how we want it to be.

Any thoughtīs on what is causing radiosity to not kick in properly in the final render?

Michael

http://forums.newtek.com/attachment.php?attachmentid=123989&d=1409599295

123989

RebelHill
09-01-2014, 01:28 PM
Are those rocks instances?

prometheus
09-01-2014, 01:41 PM
Are those rocks instances?

my first sculptris doodle made from a sculptris base plane in 10 seconds only witthout any proper idea on where I was going, ended up painting in sculptris too with a nice rock texture that is free, so much easier than zbrush to paint and save out to obj and then get it right in with uv maps and textures when importing to lightwave ..it feels more inituive and faster in that regards compared to preparing painting in zbrush.
so the rocks can and should look much better if I put my mind to it....and not just 10 second doodle it, but for you question ...yes It is one single rock that is instanced to 65 instances, the rock sculpted object is 80786 polys, made a groundplane of 12km and surface instanced the rocks on that.
the instances are varied in rotation and scale to give some non uniform appaerance, could probably sculpt a few more rock types and mix in too.

Still need to now why radiosity is so poor in the final render, after that I will wrestle and learn sculptris a little more and do proper rocks.

RebelHill
09-01-2014, 02:59 PM
Hmm... Id wondered if perhaps the instances weren't bouncing light in F9 but were in vpr... quick test here though seems to show that they do. Looked for a setting that might have caused such for them, cant see any... probably a dead end speculation there.

prometheus
09-01-2014, 03:13 PM
Hmm... Id wondered if perhaps the instances weren't bouncing light in F9 but were in vpr... quick test here though seems to show that they do. Looked for a setting that might have caused such for them, cant see any... probably a dead end speculation there.

it probably is, will have to isolate it and remove stuff, or rebuild from scratch with simpler things to verify VPR vs final render radiosity, what I noticed... the ground plane has a diffuse setting of 75% and that is affecting the radiosity lighting upon the dante ship quite a lot, if I lower that diffuse setting, it will start to match up against final render, but it still looks better and more bounced lit so to speak in VPR, so raising the diffuse much higher doesnīt make any difference in final render, only vpr.
weird, I initially thought it was a color space thing, but it doesnīt seem that way.

I can only speculate that something is not properly set up or working regarding how the rays from diffuse material are bounced in VPR vs final renders.

pinkmouse
09-01-2014, 04:22 PM
I've never found VPR to be that good in rendering radiosity, it always seems to need the bounces turned up to get any kind of match to F9

spherical
09-01-2014, 04:44 PM
Problem is, he's seeing the opposite.

gerry_g
09-01-2014, 04:52 PM
personally I would dump the dome light and use an HDR reference image for light it will fill the scene a lot better (can still use your BG image with it) I would think the mismatch is due to VPR not reading stuff correctly had trouble with VPR miss reading clamped fall off 2 with very low values on area lights recently, also think using sRGB space with HDR is way way better, lot less harsh, lot of my old HDR scenes were set to the liner default (meaning no applied setting at all) and they always rendered way too contrasty

pinkmouse
09-01-2014, 06:03 PM
Problem is, he's seeing the opposite.

Indeed, I phrased my reply very badly. :)

What I tried, and failed to communicate, is that I've seen the darker F9 renders as well, but as you increase bounces and it lightens up the F9, VPR doesn't lighten up so quickly, so it tends to a mean. I suspect the VPR GI approximation is tilted towards a few more bounces than are in the default render settings.

I think that explains things a little better...

Sensei
09-01-2014, 07:19 PM
Show rather Global Options panel..

prometheus
09-01-2014, 11:27 PM
Problem is, he's seeing the opposite.

Exactly.


personally I would dump the dome light and use an HDR reference image for light it will fill the scene a lot better (can still use your BG image with it) I would think the mismatch is due to VPR not reading stuff correctly had trouble with VPR miss reading clamped fall off 2 with very low values on area lights recently, also think using sRGB space with HDR is way way better, lot less harsh, lot of my old HDR scenes were set to the liner default (meaning no applied setting at all) and they always rendered way too contrasty

mm..maybe, but thatīs a later question..this was merely a mockup test, so I wasnīt aiming for the best lighting for a starter, but I donīt think I would split up having different light image for background and for lighting the scene, for exterior I prefer to use dp sunsky_skylight_skSun and that seems to work decently with image backdrops to if you manually match the light towards how the backdrop image look like, the image was here mapped cylindrical in textured environment, I also have a transparent large poly plane above the scene and a procedural texture in that transparency channel..just to add faked raycast shadows on the ground and the rocks so the image is more dramatic with more shadows and light as it is in the real world with some clouds above.
In this case I was just lazy with the lighting.
That poly plane isnīt interfering with the radiosity issue though as I have tested.

I am not sure what I like about the sRGB mode instead, personally I am not fond of working in that mode..since colors seems way washed out and un natural for exterior shots, no tone mapping, and I would have to rethink how I light the scenes in regards to that, I prefer the more contrasted depth than to washed out sRGB, but I will have to try working with that from scratch someday..but it seems off when working with dp_sunsky and hypervoxels sometimes.
for interior and other stuff..sure, in sRGB I feel I always have to go back and post process it, or add lightwave image filter corrections in the render, so I prefer to get direct feedback in VPR and the first renders.
Besides..if the radiosity were working as in VPR the too strong contrast in linear mode isnīt so obvious since the radiosity really takes care of most of that and bounce fill the scene very nicely.





Show rather Global Options panel..

I will check that later...busy all day with other stuff..thanks.


Radiosity I think should behave the same in VPR and final render, otherwise..drop VPR, I found the VPR radiosity looking and behaving as to be expected and very similar to Vue global illumination, the final render is not what I would like to see..the radiosity
simply donīt disperse or creep in to the object as it should.
And as I mentioned before, the ground plane with a diffuse setting at 75 is making the space ship from shrox receiving more rays it seems, lowering it to 30 or so and it will look more like the final render, but the diffuse changes has no effect upon the
final render radiosity.


Funny how you find yourself mocking around..I should do this with some sort of plan and do it properly, but It is kind of a sculptris and ligthwave doodle where you just start from nowhere without any purpose really and you start to add in stuff..
so I was just basicly testing sculptris..and since it turned out to be a rock thingy, I had to throw it in to lightwave..and since it was there I had to make some instances..and along with that comes a skydome and etc...etc..:)
Maybe I should just drop it as a project, it might serve as a previz for upcoming stuff..rocks should be sculpted with the intention of being good rocks and painted properly too, 10 seconds fiddle doesnīt cut it:)

prometheus
09-01-2014, 11:51 PM
I think I solved it, my domelight for the overcaster sky was set to backdrop, I was messing with that, should have learned not to mess with too much stuff:)
Once I turned that off, the radiosity looked better and more like the vpr, not quite exactly but it is getting there....I hope to show some samples for you guys later
to compare and in relation to having a ground surface with no diffuse compared to higher diffuse settings.

Danner
09-02-2014, 01:37 AM
When lighting a scene, to have VPR aproximate the final render as much as possible, lower radiosity settings (leave 3 bounces, but lower the number of primary and secondary rays and up the min pixel spacing) , then turn "draft mode" off. This forces VPR into doing more bounces and shadowed areas look brighter and closer to the final render (oppossite of what was happening here, but useful in most other scenarios). After the amount of light looks right I go back to draft mode for faster previews and higher final render settings for clean GI.

Sensei
09-02-2014, 05:07 AM
Radiosity I think should behave the same in VPR and final render, otherwise..

It can't be.

VPR can't do real pre-calculations.

Otherwise you would have wait for long time for even starting previewing..

How much seconds F9 render is spending at stage precalculation of GI, when you have cached GI turned on.. ?

OFF
09-02-2014, 06:04 AM
Still need to now why radiosity is so poor in the final render, after that I will wrestle and learn sculptris a little more and do proper rocks.
Color space workflow?..