PDA

View Full Version : ?? VPR 'correct', renders wrong (!!)



jeric_synergy
04-03-2013, 05:41 PM
So I achieved the correct look in VPR, but when I rendered it, the renders didn't reflect the VPR display:

113279

Baffled.

jeric_synergy
04-03-2013, 07:01 PM
Here's a cut-down version of the scene. 113280

The VPR renders 'correctly', i.e. as I expected. But the render does not. I can make the numbers/letters visible by making the surface luminous, but even with a dedicated spotlight shining on them they don't show up with a simply Diffuse surface.

113281

:bangwall: :stumped: :chicken:

Dexter2999
04-03-2013, 07:34 PM
113283
Okay, tell me if I am seeing this right. The BACK of your dial model is a single round poly just larger than the round poly with the numbers on it. BUT It is actually ONTOP of the face but since it is single sided you can see through it. This appears to be messing with the final render (as it should) but it is being ignored by VPR.

The pic is when I selected and started to deconstruct your model. It shows crossing normals. They should be pointed in opposite directions?

I think if you spin your model around you will see where the holes you modeled in the dial pass through the back and that is because the back is in the wrong location?

jeric_synergy
04-03-2013, 08:02 PM
OK, lessee: I'm pretty sure your analysis is correct, but:

why is the poly w/the normal facing away blocking the render?? I'd think it should be invisible.

The label disk is supposed to be slightly offset from the dial, but that may have gotten screwed up somewhere along the trail.

(Ugh: modeling the back fo that thing is not going to be fun: normally I'd just mirror and flatten, but... well, we'll see.)

Thanks for looking! :thumbsup:

Dexter2999
04-03-2013, 08:23 PM
Just guessing here. IT IS INVISIBLE in that you can see through it from the back. But the poly is in effect working like a two-way mirror (which work like tinted windows). Light passes through the back and hits the face of the dial. The bounced light should bounce back and hit the camera plane. However the "face" of the gold back poly isn't invisible so it is cutting the light.

dwburman
04-03-2013, 08:31 PM
Did you update the download? VPR and the render match here.

113284

RudySchneider
04-03-2013, 09:00 PM
dwburman is right! The upper-left window of your screen shot in post #2 isn't VPR.

113285

But, if you make your rotary dial and numbers two different layers, you can set the light exclude the dial and only affect the numbers, as below.

113286

jeric_synergy
04-04-2013, 12:32 AM
Dang. I need some sleep.

eKalb
04-04-2013, 11:09 AM
Do you even need the back face of the dial?

jeric_synergy
04-04-2013, 11:40 AM
Do you even need the back face of the dial?
Probably not, but it's good to be seamless sometimes if the polycount can be handled. Sometimes it saves tracking down obscure render issues.

Here's the remodeled dial, and this time I managed to actually use VPR correctly:
113298

I'm happy to say that modeling the back was NOT as onerous as I'd feared, because the new ALIGN tool greatly eased flipping around to the axes and back, so I was able to squeeze the dial down. There's still a scaling operation I'm not quite getting, but I'll address this in another thread, because really it seems only one question gets answered per thread. 8~

113299

Dexter2999
04-04-2013, 11:41 AM
But, if you make your rotary dial and numbers two different layers


And actually they should be two layers, because the numbers and letters were printed on the face of the phone, the dial with the holes would spin. So they were two separate pieces. As it is, this model would spin the numbers with the dial.

All of the sudden, I am feeling very very old.


Aaaaand you fixed it anyway and posted it while I was typing.

jeric_synergy
04-04-2013, 11:48 AM
And actually they should be two layers, because the numbers and letters were printed on the face of the phone, the dial with the holes would spin. So they were two separate pieces. As it is, this model would spin the numbers with the dial.

All of the sudden, I am feeling very very old.
Ahhh, young padawan, you underestimate the lengths I'll go to do keep everything in one layer!

In this case I use an obscure trick: assigning a weight map to the dial and constructing a skelegon (thank you SPIKY) to make a one-bone rig.

WHY would I do this? To avoid having a complicated motion on the dial in Layout. TMK, I'd have to make a null, parent, tilt, blah effing blah to spin the dial on its axis if I were doing it the usual way. That means if the object were loaded into a new scene I'd have to re-rig it (BTW, how do people handle that management issue??? In the past I've made 'loader' scenes and used LOAD FROM).

By using the skelegon method I get two benefits: The object is more portable, and the skelegon can be easily aligned with the axis. I just have to remember to use LOCAL COORD SYSTEM when I'm animating it.

JoePoe
04-04-2013, 12:33 PM
Yes, it was the back side poly with no holes.
Because....... even though it's single sided with the normal facing away from the camera, and in effect "invisible", it will still cast a shadow!! Hence no numbers. :)

And, you must close the back (which you have already done), otherwise the shadows will be all...... screwy.

JoePoe
04-04-2013, 01:00 PM
Jeric, I've attached the model with my version of the back sealed (I also took the liberty of changing your smoothing angle... I think it helps). I can go into the method if you want. The way I went about it the angle of the model is irrelevant.

I don't have Viper.... this is shaded solid in Modeler.

113309

jeric_synergy
04-04-2013, 01:11 PM
I can go into the method if you want. The way I went about it the angle of the model is irrelevant.
Please do!

JoePoe
04-04-2013, 03:50 PM
1) CUT/paste bottom plane in new layer
2) COPY/paste number hole walls in another layer
3) Boolean stencil holes into plane, delete circles
4) CUT/paste back into original and merge points (I knew that there were 288 points to merge. Automatic setting didn't get 'em all, so I used fixed distance. 70um did the trick.)

(didn't give me a problem, but sometimes a poly with that many points may get smoothing issues. Just multishift in a tad.)