Two Sided Patches

[Rodney]

A few years ago Robert posted an example of texturing of a patch where opposite sides show different materials, textures, etc.

I'm not sure if we every found out his approach but for a little while I thought I had it figured out and did some exploration of my own.

 

Here's a very basic example that colors opposing sides of a patch red and blue.

The dangling splines are there to provide some curvature on a single patch which would otherwise be straight/linear.

 

 

 

[robcat2075]

That's pretty good.

[Rodney]

Here that patch as Shaggy's cape.

 

I've run into a problem with images/reflections on the patch and I'm still investigating that.

Straight color works fine in the reflection as seen in the attached.

[Rodney]

When I try to apply an image I get some odd stuff.

In the attached there is an image applied to red side. No image on blue.

The reflection appears to be of the color's inverse (?).

[Rodney]

Removing the image the proper reflections show up... (please excuse Shaggy's exhibitionism.. there's a pass through of his cape/butt)

 

At least I've got that side of the issue isolated.

[robcat2075]

The reflection thing is weird. Try an image with an image in it, something other than a flat color.

[Rodney]

Try an image with an image in it, something other than a flat color.

 

That's what it was. It was an image of a Disney character.

I was planning on trying a flat color next to see what would happen..

[Rodney]

Ha! User error.

A quick trip to the Tech Reference got me headed in the right direction.

[Rodney]

Here's a couple renderings working as it should.

[Rodney]

It should be noted...

 

It'd probably be easier to create the same/similar effect with two meshes slightly offset.

[robcat2075]

I dimly recall that my technique was a compositing process.

 

It used the fact that the "wrong" side of patches with SSS on them would render as black to create a matte to composite two renders together.

[Rodney]

My memory recalls that it had something to do with compositing so that makes sense.

Yours was one of those instances where by posting your results and THINKiNG that I knew your approach I explored a completely different approach.

 

And the answer to that other approach is: Multiple Light Projections (with or without image gels (rotoscopes)

 

There are a few 'tricks' that have to be used in order for the light not to bleed to the other side of the mesh as well as a caviet on use of projected images.

For relatively flat meshes the process is rather straightforward:

 

(All within the same model)

Create mesh

Place one light (klieg) on each side of the mesh.

Adjust as appropriate (to maintain a mesh that is as flat as possible).

(Then make all mesh curvature adjustments in an Action... I haven't tested poses but those should work)

 

 

The lights can be placed outside the model but in many cases that defeats the purpose of adjusting the mesh (in an Action) while maintaining the two sided projections.

This could be done via constraints but when the lights are inside the model they get automagically constrained for us.

 

A second trick (more often than not) is required to keep other models in a Chor from receiving the projections on our target mesh and that is the use of at least one Light List.

This takes advantage of the fact that when a light list is first created we have the option to exclude other lights from models without light lists.

 

A specific downside to the approach is that meshes cannot be overly curved without risking bleed from the other projection.

A possible solution to that would be to add additional lights to account for those specific areas.

 

Another downside then is that the target mesh in most cases will need to be a separate model.

In order to be of much use in animation this will require constraining that mesh to another object in the Chor.

 

In most cases the model itself should not be adjusted but rather adjusted in Action or Chor as this will maintain proper light projections to maximum extent.

In other words, the adjusted mesh doesn't lose the projections as easily when the original model maintains its flatter (or linear) curvature.

 

The issue I was having with the image reflections related to trying to add the image directly to the light rather than the intance of the light within the Model in the Chor. The image appears to be applied correctly to the light but it isn't. The Tech Ref got me back on track with regard to Right Clicking on the Light and selecting New/Rotoscope. I'm not sure why we can't just drag and drop an image onto a light... but that's just not how it works.

 

And while I haven't experimented much the colors and images can also be animated.

 

Edit: An added 'gotcha' I'll have to investigate... light gels within models don't appear to save. As such they have to be re-reattached upon opening the Project again.

Ah the joys of living on the cutting edge.

[Rodney]

Here The Spirit demonstrates the technique with animated images and transparency...

sprited transparency.mov

[robcat2075]

 

 

A specific downside to the approach is that meshes cannot be overly curved without risking bleed from the other projection.

 

that's the one that's probably a deal-killer for use on cloth where you can't explicitly control the shape. Cloth will get curved and folded.

 

 

 

So far, the SSS method is the only way I know of differentiate between front and back of a single thickness surface in any shape...

 

[robcat2075]

Here I moved the light to show that it's not dependent on the environment lighting

 

[Rodney]

that's the one that's probably a deal-killer for use on cloth where you can't explicitly control the shape. Cloth will get curved and folded.

 

I specifically developed the approach with cloth in mind and (as I recall) it worked fine.

The key (with cloth) is that the original model is flat.

The cloth simulation then picks up from that.