Render 20hours?

[Ian Jones]

Sampling max 64 and the size of the render A4! sounds like your major problems.

[Saku-Pekonmaki]

GI 200
I´m usualy rendering with FG only. with couple of lights. (mr lights or photmetric)

don´t use skylight, don´t use skylight
don´t use skylight, don´t use skylight
don´t use skylight, don´t use skylight
(skylight generates GI mostly thought for radiosity scanline renders. and will not do anything in MR that a simple SPHERE around the object with gradient ramp illumination material with gloss around 10 will not do 1000 times faster)

tip 1.

outeriors ! 1 sunlight with your prefered shadow method, raytrace is good.
If u dont like sunlight, then use something else.
but stick to one light.

tip 2.

interiors ! same as above. plus the lights that uthink u need inside to create the correct light from corresponding ...hmmm....... lights?! :)

I dont use lights inside unless there is no way for the light to get in.

The photons in GI U can squeeze up to hundreds of thousands without really affecting the render time, but having high FG ratio will give you a big time difference. I usually use a FG setting around 200-300 in a final image. radius to use should be like every one says.... 10% of average model size. leaving it to default can cause artifacts depending on the scale.

A4,A3 & etc. gives a bigger pixels per inch ratio = width & height. wich is good for printing materials. but way too much for just a pic on the net or desktop.

caustics is also a heavy time killer. between 10-100 is good depenging on your scene.

samples per pixle can be tricky one if u don´t know what they mean. (not that I know :) )
MR renders out with "boxes" and the higher between value you have the longer it will finalize each renderbox. putting litterally more time for each pixel and smoothing out jagged edges. the bigger the image the bigger the pixel ratio should be if to get "GOOD" render. I use 4/16 ratio and the pictures r just fine. higher ratio I believe is just waste of time, but that depends on what you are doing (ofcourse). also the difference on using "BOX" or "mitchell" smoothing will make a time difference on the render.

uff.. then there is alot of scene tweaks.

My advice is to have alot of memory.

and ofcourse Guys feel free to crap me down. :)

[digones]

here's my two cents:

try dirtmap :) oorrrrr the default occlusion shader that comes with Maya 6.5 to generate your occlusion pass and post composite in photoshop...

Saku: Hey, man, thanks for the tips! :applause:


cheers

[Dascrewup]

hi

thanx guys but saku i did not understand what u mean by this?

"skylight generates GI mostly thought for radiosity scanline renders. and will not do anything in MR that a simple SPHERE around the object with gradient ramp illumination material with gloss around 10 will not do 1000 times faster"

do you mean that i make a sphere and insert a omni light inside it? and what is gradientr
ramp illumination?

Thanx for the help guys :)

[TheWriter]

I agree sampling can be a big time-rate killer. For final renders I set min to 1 and max to 4. Anything higher than that and I can not see the difference. There seem to be a few different AA algorithms to chose from, I stick with the default though wish I knew a little bit more about them.

[cpan]

For final renders I set min to 1 and max to 4.

holly sh*!!
DON'T set AA to 1, 4 but maybe 0, 2 but lower the contrast to 0.05,0.05,0.05 or lower to get better results.

In Render Globals there's also a per object AA that you may want to use in order to lower rendertime. Also 3000 photons... maybe 300.000 photons... 3000 photons aren't accurate at all but fast... perhaps you should use multibounce instead of GI photons, they offer A LOT better quality and use fewer RAM

[TheWriter]

Are you using max? I do not think 0,2 is possible, as my interface only lets me use fractions.

[cpan]

oh, i thought this is a maya thread :eek:
in maya & mray standalone 0-2 means 1-16, see szalbocs (http://www.impresszio.hu/szabolcs/MentalRay/MentalRaySampling.htm)'s explanation of AA. So set your samples to 1/4, 16 (-1,2 in maya) instead of 1/4, 64 wich you should use only when glossy objects are needed. (eg. DGS shader)

as for the photons, are you getting warning such 'no photons stored after blablabla'? If so, perhaps you didn't attach photon shaders to some of your mray materials. Also for caustics, it's very important to point your light to your caustic-caster objects, else you'll get slow renderings and warnings.

BTW. from what i'm remembering max's mray was always very slow at emitting photons, dunno why, so you may want to turn off GI&Caustics and leave only FG... but you won't get good indirect ilumination unless you turn on FG multibounce and set the bounces to 3-6.

[TheWriter]

That page was down last night but just read through it today. Interesting, it claims the box algorithm is the worst, yet Max has it set as the default system.

Also it seems some of the other AA methods claim to produce sharp anti aliasing on edges WITHOUT blurring. Sounds impossible to me but hell, I'll give it a shot.

[cpan]

yes box filter is the fastest but produces bad AA and Mitchell is the best, I allways use it.
here's another great place to learn mray: LA mray user group (http://www.lamrug.org/)

[TheWriter]

Last night I switched to Mitchell, and to be honest I think the aliasing looks like shit. I had setting on 1,4 and this is my result? In this case, it looks as though I might as well stick to Box.

Maybe i'll turn on jitter tonight and see if that helps, but I doubt it much. It's just.... crap.

Mitchell (http://www.highpoly3d.com/writer/wip.png)

[Ian Jones]

1,4 is not really going to get you great results on box or mitchell. Up to 1, 16 or 4, 16 and you'll see the difference.

[cpan]

@TheWriter
well, if you choose Mitchell you must set both the filter widh and filter height to 4,4 or higher
since this filer takes more pixels in account than the others.
then lower the AA contrast to 0.05, 0.05, 0.05 or lower for better AA cuz this controls the Adaptive Sampling. If you don't need Alpha, then set the Alpha Contrast to 1.
As for the sampling, for the final pic you may want to set it to 0-2(maya) and 1-16(max)

[TheWriter]

Well my scene finished just recently. I jacked it up to 1,16 and much better improvement. However I do not see anything special, it looks as though as it's the same as 1,4 for Box mode. I'll try 2,16 on next pass and hopefuly it doesnt take 2 days to render.

BTW, there is a Lanczos method as well under Mitchell, what is the latest word on that?

As for some control stats, default is already set for 4, 4, on filter size for Mitchell in Max.
Also the contrats are 0.051 and 0.05 for alpha, which is Max's default.

[johnmoney]

20hours?????

[Saku-Pekonmaki]

Dascrewup (http://www.cgtalk.com/member.php?u=104528)!

what I mean is to make a sphere around the object.
in diffuse color/material u make gradient ramp for an example (to make a sky) starting with a dark blue color, then close to the left edge of the ramp a lighter blue color, and then white.

with this u get a nice blue color illumination on your scene, but only if your FG is turned on.

the color should always represent the "ambient" of your scene.

sorry for my english :)

as for rendering, highpolymodels will not affect the actual render time much, but will suck up alot of memory. wich can cause the render to crash or hangup.

[TheWriter]

Ok an update. I rendered out both a 1-16 and 4-16 AA on my scene. I then put both images in layers in photoshop then swaped between them to view any differences. I noticed no change at all on edges. However, interesting to note, it seemed as though some flat areas in the image were less grainy with 4-16 AA. Strange, as I thought anti-aliasing only was supposed to affect edges of polys, not their surfaces? It was a very slight change mind you, but noticable in photoshop.

This of course, was still using the Mitchell process. And I am still not satisfied with the result.

[Saku-Pekonmaki]

u should post the images :)

[TheWriter]

Bah, not much to see. I'll see about working more on my scene and in the next few weeks i'll get back to the aliasing issue again before I send off my final print.

[Cuprum]

Two filters that seem to be the best are Mitchell-Netravali and Catmull-Rom

[TheWriter]

Hmm, I am familiar with Catmull-Clark sub-divisions, but the Rom is a new term. Oh well, I see that Mitchell is quoted a bit in places, but I just do not see much improvement with it.

I think I may spend the next two weeks doing a lot of Final Gather tests.

[TheWriter]

Well ok something worth posting (cause it's wrong). In case no one in the max threads figures it out... I used a dome and tried to get the material to emit light in MR. Funy thing is only materials with blinn shading are reflecting the light. Puzzling...

http://www.highpoly3d.com/writer/wip/monopoly/light_test_1.jpg

[TheWriter]

Ok this dome stuff just doesnt work out for me in MR. I think I should just swtich back to the simple and old Skylight method? Tried it out, though obviously it needs some tweaking. Just been ages since I've used skylight so I almost forget how to clean this up. I'll crank up the FG samples and hope that does the trick. Suggestions?

http://www.highpoly3d.com/writer/wip/monopoly/light_test_2.jpg

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.