Shadow Catchers

[Richard Hebert]

This is a feature found in Poser 7. Geometry can be instructed to catch shadows but when that part of the scene is rendered only the shadow gets rendered allowing for the shadow to be composited in live video in another program. Not sure if I'm being real clear on the concept... but does anyone know how to achieve something similar in LW? I need the shadows that the geometry casts but not the object receiving the shadow.

[RollerJesus]

Sure can do.

In Layout, select the object you want to not be a "shadow catcher". Hit "p" to bring up the properties window and meander over to the "Render" tab.

Towards the bottom you can toggle the object to:
"Self Shadow" - Cast/Receive on itself
"Cast Shadow" - Be a shadow pitcher
"Receive Shadow" - Be a shadow catcher

[Sekhar]

I think he means just the shadow. Richard, for the object that needs to catch the shadow set its surface to create alpha from shadow density - go to Advanced in Surface Editor, and for Alpha Channel select Shadow Density. Then, when you render you'll get an alpha that you can use to composite. You might want to set other objects as unseen by camera so they don't appear in the alpha.

[RollerJesus]

Apologies, should have read the post more thoroughly.

[Nangleator]

If compositing inside LW, then surface the shadow catcher with "Front Projection" mapping. Carefully balance it's Diffuse and Luminosity settings so the background image looks good on it, while still showing a shadow.

If compositing outside LW, well... Run a pass so that all foreground objects are black, the shadow catcher is white, and the shadow appears on the shadow catcher. Then multiply that layer over the b/g plate, and add the full color animation on top. Is that too complicated? I don't composite outside LW.

[Richard Hebert]

Thanks for the replies, this gives me a good basis to start experimenting. Every response has been very helpful.

[Richard Hebert]

OK, I've turned off objects casting shadows (unseen by camera) and I have a shadow on my geometry... In this case a road. What I want I want to do now is make my model road invisible to the camera and still see a shadow for compositing. When I make the road invisible to the camera, my shadow disappears with it. Any way to keep the shadow or am I missing a step in between? Sorry for the headaches.

[UnCommonGrafx]

Save the image in a 32bit image and it will you will be able to make this object go away during compositing.
Setting to make Model Road invisible:

[Sekhar]

This is a two-step process: you need to capture the alpha (that corresponds to the shadow) and use that to composite. See http://www.newtek.com/forums/showthread.php?t=59612 for details on this - I'd posted a sample you can download and check out (use the second Shadow.zip).

[Richard Hebert]

OK, it's working for me. Thanks guys for all the input. You shortened the learning curve yet again!

Richard

[Richard Hebert]

Hi Again Everybody,

My problem with 'shadow catching' persists. I thought the problem was solved earlier but then realized that the shadows were dense (opaque) and so the issue went unnoticed. The surface catching the shadow is gone except where the shadow falls. When the shadow has varying density, the underlying object is still there. I need to have just the shadow with its varied density alpha. Hiding the surface catching the shadow (unseen by camera) eliminates the shadow also (regardless of saving as .tif 32 file). Unfortunately, I need this function to work if possible. This is an animation project with .tif frames. Could it be that the software cannot handle this particular task yet? Thanks in advance.

[Surrealist.]

I think you missed a step. Go back to what Nangleator said. If you make the road object white for your shadow pass I think it will give you what you want. Unless I missed something. This will give you your gray scale shadow that you can composite using the alpha map you render separately in hthe output tab.

[faulknermano]

robert, are you trying Sekhar's method? this is the path you want to go. but you must realise that the final output is a black image with an alpha channel that corresponds to the shadow density.

you never hide the shadow-catcher (e.g. the surface that catches the shadow). only the shadow-casters are hidden from camera but able to cast.

[Richard Hebert]

Ok, let me see if I have this correct. The shadow density setting doesn't really give me just an alpha channel with varying density but rather gives me the shadowed surface of the object. This is cumbersome at best since Poser being a much less sophisticated program actually gives a true varying density alpha not just a shadowed surface. This seems like an area that LW may want to improve in a later release. Will make compositing with video less cumbersome. Final compositing is not really an option for the footage that I'm having to use. It would send my render times through the roof if I had to do it that way and grain matching would not be as easy either. Thanks for help.

[Sekhar]

You're probably picking up the rendered image rather than the alpha. In the render window, use the pull-down menu at the top right to select alpha instead of image. If you make all other objects unseen by camera, the issue of render time shouldn't come up because nothing is going to get rendered other than the catching object.

If you would rather have the alpha generated as the final rendered image, make all your lights black with white shadow (again, with all other objects unseen by camera). The final render will be a black image with white corresponding to the shadow on the catcher. If you render this out as an animation, you can use either the final rendered black/white image as your luma key or use the embedded alpha as your alpha key - they give slightly different results, choose whichever you like.

[Surrealist.]

Well with the attached scene example what you get is an alpha that will match the density of the shadow - edges only. It gives you the ability to then composite your shadow whatever graininess, color or density it is. That you can make on a separate pass by rendering your shadow onto a white surface. This will effectively give you the ability to have a layer with the shadow only and composite that with the rest of the image. You would then use the opacity of that layer to mix it in with the surface.

What I do if I am having trouble with something in a scene is I close it, do some basic tests until I understand how it works, then open up my scene and apply the information I have learned from my tests.

EDIT just saw the other post...

[Richard Hebert]

When the image is rendered, I'm selecting Image>Alpha from the dropdown menu and exporting that as a .tif32. When I bring the resulting image into After Effects I have no alpha... just a B&W RGB file. Is this supposed to be?

[Sekhar]

Yes, if you select alpha from the drop-down, what you'll get is a b/w image that represents the alpha (not an image with this info in the alpha channel). You could use the entire image as the luma key in AE.

If you save the full image itself (rather selecting the alpha), but save it along with alpha channel as an image, you can use the alpha channel from that image for an alpha key in AE.

I made a simple example to show all this. Attached is a cube on a floor (the shadow "catcher"). The floor has the shadow density to alpha setting we talked about. After the render, I saved a Shadow.exr for the image+alpha channel (with "image" in the drop-down and saving EXR with alpha). I also saved just the alpha as an image (by selecting "alpha" in the drop-down) as Shadow-Alpha-Only.exr.

In AE, Comp1 uses the alpha channel of Shadow.exr as alpha key; and Comp2 uses the entire Shadow-Alpha-Only.exr image as luma key. Both give similar (though slightly different) results. These should give you the different options. You could also try the black light/white shadow thing I wrote about to experiment.

[Richard Hebert]

OK, I've got it now. Seems a little cumbersome with all the CG animation that I may wind up having to do but I can live with it since the renders are much more realistic than what smaller software is giving me. My biggest bug-a-boo is that I'm still using this software on an older and slower (1.5 GZ) Single Core Mac Mini. I'd really like to employ a lot of the suggestions being offered on the forum but my computer just can't handle the load so that's why I'm looking for ways around what would normally be considered the 'right way'. Making two or three pass renders is technically correct but my projects wouldn't be done until this time next year if I did everything that way. Thanks for taking the time and helping through this process. And yes... I will be upgrading my computer to handle CG projects better. Just have to figure out which banks to rob... that's a ha ha for all you FBI types!

[Richard Hebert]

This is a little off topic but is there a box to tick to see OpenGL shadows? Can't tell shadow direction when I move the light source.

[Surrealist.]

To your point about this being cumbersome. It is. But it is a lot faster than tweaking and rendering again. Also this is why 64 bit and larger ram was such a big deal. Because now things you would composite before because of RAM limitations can now be done in one pass.

Here is an article on current techniques for compositing (http://www.spinquad.com/articles) that may help.

[UnCommonGrafx]

Richard,
I believe you are seeing the task of multipass rendering as tedious as opposed to functional. Allow me a bit of musing...

If you have everything in one pass, it may take five mins per frame to render out.
What's the equivalency if you render out these passes you need?

Ok. Now, just for the math of it, and after looking at the above article, realize that you have to recolor, re-arrange, re-something. Is it quicker to re-render the animation or to tweak it in your compositor?
How's that computer with your compositing program?

MultiPass is really sweet in an under-powered environment. It's highly under-valued, as well. Organization is the hardest thing about it...

[faulknermano]

i think that even with the absence of a multi-pass 'helper' like http://www.lwpassport.com it is still bearable enough as long as you have a system. it's certainly a software issue, which may mean intrinsically lightwave, or 3rd-party tools, but methods may be designed so as to make the next time less of a hassle.

i have my own break-out utility, but before i did, i used to write on Post-It notes the settings and reminders needed for shadow-catching, shadow-casting, occlusion, etc. i would save out .srf files of shadowcatchers and occlusion objects. and before breaking out passes manually i write down how the scene is going to composite in layers.

so what i'm trying to say is that methods and rules and guidelines can be designed and followed giving reliable and less tedious results. for a lot of the tediousness goes into the thinking of which parameter is switched on or off. we leave that part of the thinking to the system.

[Richard Hebert]

I'm seeing an issue with reflections not rendering over the alpha channel as well. I have glass in the shot that has reflections on it but the reflections aren't visible visible where there is alpha. Is this the same issue as the shadows? If so, how does anyone composite over live video outside of LightWave? I need to have this project composited in After Effects. Once again, if Poser has these issues resolved how did NewTek miss the boat? I just hope I'm doing something incredibly wrong. As far as multi-pass rendering goes, I just don't have the time on this particular project to handle all of the facets that goes into the process. That will come at a later date. Thanks for all the info. on multi-pass though. It will come in handy on future projects.

[oobievision]

Yes, if you select alpha from the drop-down, what you'll get is a b/w image that represents the alpha (not an image with this info in the alpha channel). You could use the entire image as the luma key in AE.


Try rendering it out as a targa sequence instead. AE has an easier time translating the Alpha.

[Sekhar]

Richard, no offense but it's very hard to help you unless we know exactly what you're trying to do and how much effort you're willing to put in (and what tradeoffs you're willing to make). Why don't you post your scene (or a simplified version of it) so we know precisely what you'd like to do and how you're able to do that easily with Poser and not with LW? LW is way more powerful than Poser in general terms (I have Poser 7).

[geoff]

Sorry to hijack this thread, but I've been sat for the last couple of hours trying to work out how to have a object show "only" the shadows cast on it at render time ( no comping involved ).. I.E white background, and coloured sphere and a white ground object but only the sphere and shadows cast on the ground object by the sphere are shown when rendered, everything else is white. I'm sure I have done this before but for the life of me I can't remember what I've missed. I've got brain ache at the moment and would appriciate it if somebody could help out.

[faulknermano]

geoff: my recommendation to you is to notch up your ground plane's diffuse by a greater amount, like 200% (or if extreme light angles, greater amounts like 8000%), and shut down ambient intensity to 0%. the problem comes from not having enough white on your ground plane when the light rotates at certain angles. however, this method works well if you're not working in floating-point, because a high diffuse will give you a 'super-white' plane.

[geoff]

Hi faulknermano
Yep that was the trick.
I knew it was something simple, but some days you just get brain block.

Thanks again