PDA

View Full Version : fast skin/simple skin bugs



Kuzey
02-23-2009, 05:16 AM
Hi guys, there is something weird going on when using fast skin or simple skin nodes.

First, I'll start from the beginning. I used a shader for the sofa that looked great in LW7.5 but was terrible in 9.6. I have no problem with that as there were major changes after 7.5 etc in the render engine and what not.

So, I swapped that for a hair texture that uses simple skin node, it looked great, took about two and a half hours to render. I then made changes to the camera and object locations/composition etc. so there was slightly more of the sofa in the camera view than before. The "computing preliminary radiosity solution" part took about 8 hours and the render hanged at about 28 percent render several hours later.

Then I swapped the simple skin node for the fast skin node to seep up the render. I forgot to connect the material output to the material input of the surface node. The clean render is the result of that and once I connected the material to it's rightful place...I got the other almost black render :hey:

I'll try to reduce the scene to it's bare elements and up load the scene.

Kuzey

Kuzey
02-23-2009, 10:25 AM
Hey Chilton, here is the striped down scene for you to have fun with. Make sure the PC guys test it too...just in case it's a cross platform bug :hey:

Kuzey

toby
02-23-2009, 04:53 PM
If you disconnect the Function input to the Veins node it renders ok. What's that supposed to do?

Kuzey
02-24-2009, 06:03 AM
If you disconnect the Function input to the Veins node it renders ok. What's that supposed to do?


I have no idea, just playing around and connecting things on a whim as I usually do. :D

The original simple skin version looked great on the render I managed to get out, in that scene there were a couple parts of the sofa still with the 7.5 texture....I failed to change.

Once I updated the old surfaces and changed the composition of the scene, it had a huge jump in render times and ended in layout freezing around 16 hours later. It's not like the whole sofa is in the camera view :D

There is something very wrong with the skin nodes, or how they interact with the render engine, or other surfaces in the scene that causes the blow out in render times and or app freeze. Maybe, there is a limit to the number of surfaces that can have the skin nodes applied to them.

Anyway......

Thanks Toby...I'll play around with it some more :thumbsup:

Kuzey

toby
02-24-2009, 12:01 PM
I see now, I guess it's a way to alter the procedural calculation? You get a solid color if you enter a single value, seems like a procedural would be a useful thing to plug in instead. Maybe try a different one?

Kuzey
02-24-2009, 02:56 PM
Thanks again, I removed the connection to the function input and the render still turned out bad...but not as bad as the first one though :D

So...some other settings might come into play as well.

Thanks for the tip...I'll try that and see how it turns out :thumbsup:

Chilton...have you looked at it. I'm thinking no matter how silly I might be in the node editor, it still shouldn't result in a crappy render like that :hey:

Kuzey

toby
02-24-2009, 04:46 PM
I'm trying to re-create this in a barebones way, with only the elements that cause the error; for example, turning off caustics and all raytracing does not fix the error in the original scene, turning off radiosity interpolation or removing the function input does. I started this in a new scene, but I can't get it to error so far -

toby
02-24-2009, 05:36 PM
Ok I found it
It's because you have the procedural bump going into the function, and then the Alpha from that node plugged into the Refraction Index of the skin shader, it's repeatable. But I don't think it's a bug, or maybe it's a low-priority; refraction index is totally unpredictable when you plug in random values like procedurals. I've seen glass make scenes go black because the camera thinks that everything *outside* the glass has refraction. This one looks like it's subtracting a lot of value from the groundplane around it, cancelling out the lighting, or returning a non-color value or something whacky. Pull the camera back and you start to see the ground further from the object.

Kuzey
02-25-2009, 06:19 AM
Toby....you're a super detective :thumbsup:

I'm on Xp on the moment so I'll play with it later. I hope Newtek does take a second look at the skin nodes, I'm sure they can do with some optimizations and what not.

Where is the Newtek Node expert :D

Thanks again!

Kuzey

toby
03-02-2009, 11:27 PM
Was reading up on nodes, check this out :

One very advantageous rule is that when connecting
dissimilar types the output type will be automatically
converted to the input type that you are connecting it to,
in the most intelligent way possible. There are only three
exceptions to this and they are:

1. The Normal input in the destination Surface node
requires the incoming type be a vector.

2. The Bump input in the destination Surface node
requires the incoming type be a vector.

3. Connecting another type of output to a Function input
will not convert it to a function type. The connection
will be made and some of the information from the
non-function output may be used but the results will be
unpredictable. Functions are the only type of connection
that are actually bidirectional. This bidirectionality is
transparent to the user but Function nodes receive
information from the node they are modifying, alter the
information in some way and then send the results back
to the node they are connecting to.


Weird! I can't imagine what info going backwards in the nodes would do :confused: but maybe that's what you're seeing!
It sounds like what you had set up with the function gives unpredictable results, piping that into refraction would make it even more unpredictable.

Kuzey
03-03-2009, 05:26 AM
How did you find that...the 9.6 manuals or some other source :thumbsup:

There really should be an error message of some kind instead of allowing the connection to go ahead.


Kuzey

toby
03-03-2009, 02:53 PM
Yes it's in the 9.6 Surfacing and rendering chapter. You can download that chapter by itself from your account page.

So I guess what it's not really 'info going backwards', it just works by reading the procedural you have it plugged into.

Kuzey
03-04-2009, 04:31 AM
Thanks Toby,

Here's an updated render...it's getting there, but now I need to work out how to get specularity and maybe a bit of reflection etc to show up. I think some of other input connections might be affecting those settings....so, more experiments to come :hey:

ps. I added a scalar node and plugged that into the function input and also the refraction index input.

Kuzey

toby
03-04-2009, 01:44 PM
Wow that's very nice!

zardoz
03-17-2009, 04:11 AM
Hi Kuzey,

The glove's surface...is it a material node or 'normal' stuff?
tx
LL

Kuzey
03-17-2009, 07:39 AM
Hi Kuzey,

The glove's surface...is it a material node or 'normal' stuff?
tx
LL

A combination of nodes and the old surfacing system....here it is, there is also two unconnected nodes I left in there as well :D

Kuzey

zardoz
03-17-2009, 08:14 AM
tx for sharing the preset...I have to create a plastic ball like the one in the hamster in Bolt and your preset looked like a good start.