Juraj's Renderings thread

[ASaarnak]

Nice new images Juraj! :)

"Or too much use of 3dsMax composite node ?"
I used a composite node with a few rather huge textures for my floor in my "Warehouse Conversion" project, after I removed those the render time were something like 1/4 of the original image.
I don´t remember in detail everything since it was over 1 year ago I did that project. I can of course open up the project and check.

I´ve kinda thought of a connection between crazy rendertimes and 3dsmax composite but I never really tested it, because "how the f* would something like this make such a difference"...I use composite extensively in all my projects. So I should be using CoronaMix´s on top of each other now instead?

[romullus]

While I suspect this could be Corona's behavior, if it's 3ds Max I give up :- )

Seriously, in Unreal4, I can use 100 nodes, and it will compile into same-speed shader as zero node 4 texture shader.
I mean, how can, a stupid node, affect actual rendering performance like that ?

Slow rendering of texmaps is definitely 3dsmax fault. Each texmap is its own microcosmos, with its own naming conventions, internal layout, approaches to optimization (from "ok" to "horrible"), ... But generally it seems nobody bothered to make the nodes render at least reasonably fast. But it should not affect noise level, only speed. So rays/sample and noise/pass should stay the same, rays/second should decrease.


As for the noise: try rendering crop of the region with all CESSENTIAL passes so we can see which one contains the noise

I wonder if Corona Standalone could be solution here? Probably not now, but sometime in the future, when obstacles with porting most used max nodes will (could?) be solved.

[Ondra]

yes, we can reimplement 3dsmax nodes to make them multi-platform, faster, and bug-free. We already did it for falloff and particle maps

[Juraj]

Here is simple scene, at early stage of project. But it's already noisy, and kind of slow thus. I attach different versions I tried to pin down the issue. Without any result.

I would gladly give the scene to somebody (Maru,etc..) who could have a look.

As you see, it took 1000 passes ( 2 hours on 160 threads for 2560px, small, preview render ) to get <1perc. noise ratio.

I then started stripping the scene by isolating only the building we see (image 1) for some increase, then diffuse only except windows (always thin mat, but still multiple panes), and last diffuse only, and no windows.
I don't think it gets dramatically less noisy.
I also tried portals, maybe I am stupid, but no matter how I use them, I cannot see any difference. Can I put them inside here ?

[maru]

Hi Juraj, sure, I can check the scene. You can host it somewhere and send me the link here: https://coronarenderer.freshdesk.com/support/tickets/new or directly to miodek@corona-renderer.com
Also, is it 1.4 or some exotic daily build? :)

[tolgahan]

Hi Juraj

How do we decide about the pass number on the scene? Is there anything we should pay attention to ? Is there a huge difference

between 1000 pass and 300 pass visually ?

Thanks.

[Juraj]

Hi Juraj, sure, I can check the scene. You can host it somewhere and send me the link here: https://coronarenderer.freshdesk.com/support/tickets/new or directly to miodek@corona-renderer.com
Also, is it 1.4 or some exotic daily build? :)

Thank you, sent. 1.4.

If you give me adress I'll send you lot of beers :- ) Or if you need anything.

I'll be thankful for any kind of feedback

[Juraj]

Hi Juraj

How do we decide about the pass number on the scene? Is there anything we should pay attention to ? Is there a huge difference

between 1000 pass and 300 pass visually ?

Thanks.

Hi, we don't :- ). Usually by 100 passes (no matter what your settings were) a full AA has been completed, so only noise needs to be resolved.

Also passes depend on settings. 100 passes on default 16/2 equal 50 passes on 32/2 (GI/Directlight).

Previously, I had most scenes resolved by 100 passes, 200 at best for complicated scenes. But recently, I have lot of scenes that require 10 times more. I don't know why yet.
Better indicator is noise ratio. I have personally find I can consider scene to be finished under < 1.5perc. ratio. But if the scene has large amount of white surfaces, than it must be even lower, like <1perc.,
because the noise is very obvious in clean white surfaces.

[tolgahan]

Thanks.When ı got u online I would  ask barrage of questions :))

[Juraj]

Ask away :- ) But I will answer probably over the weekend, have a deadline tomorrow.

Your Time traveller house inspired me to learn 3D in college, so I hope to repay somehow.

[Juraj]

I am deep in deadlines, but I see some light at the end of tunnel, and finally some projects to be published.

In some spare time, I put this on web, a full 360 in high-res :- ). Btw using open-source, free Pannellum ! Yeah, say no to Krpano.

http://panorama.talcikdemovicova.com/

[Ondra]

nice, do you also have VR version?

[Juraj]

nice, do you also have VR version?

Sadly no. Don't have any kind of glasses for it yet :- ) Don't want any yet either.

[aldola]

fantastic juraj, im not a fan of panoramas but i love the detail on these, how do you achieve so detailed bump on the floor? just a map?

congrats

[Juraj]

fantastic juraj, im not a fan of panoramas but i love the detail on these, how do you achieve so detailed bump on the floor? just a map?

congrats

looking at it back now, I don't think it has a very nice bump, I used Bertrand's planks, but they were very low-res (for today standard)  so not good bump out of those, so I overlayed it with simple scratches ( I think the ones that came with cg-source ).
I am absolutely baffled why it looks inverted..

Today, I would use a good normal map made from very low filter (2 px ? ) high-pass filter of the diffuse. ( Diffuse---> High-pass----> B&W----> Contrast----->Normal map generation (Quixel, Pixplant, Bitmap2Material, anything...)