|
|
Compositing Dribble-Rendered Elements With Live Footage
About Truespace Archives
These pages are a copy of the official truespace forums prior to their removal somewhere around 2011.
They are retained here for archive purposes only.
Compositing Dribble-Rendered Elements With Live Footage // Roundtable
Post by frank // Dec 26, 2007, 10:59am
|
frank
Total Posts: 709
|
My old workflow for compositing trueSpace animations with video footage was thusly:
- Render objects/characters by themselves (no background, no floor planes) to TGA (32 Bit, alpha transparency)
- Add ground plane and proxy objects to catch shadows - this plane and other objects are all bright green (0,255,0) with ambience turned up. Paint all foreground objects/characters invisible using the Invisible shader for ShaderLab. Render.
So I would basically have two "plates": the characters and props on one, and a green-screen plate with nothing but shadows cast by the characters and props.
All that would be loaded into After Effects where it gets composited over the live-action footage. The "green-screen" layer, which gets chroma-keyed, sits in the middle of the 3D elements and real footage, and is then desaturated to ensure no green fringing occurs, with further blurring and opacity changes to add to the realism. This of course could all be done without affecting the 3D/character layer, which is why I don't just render characters AND shadows out to green in one pass.
That gets composited, precomped (basically "flattened" together), and then color-corrected for the final render from After Effects.
SO.... that brings me to this:
I am really wanting to composite some motion-tracked footage with 3D elements rendered by Dribble. Getting the 3D elements rendered to TGA or PNG yields the transparency/alpha map I want, but the shadow-catching part is the difficulty. Here are the things I've tried:
- Various ways of shading the objects invisible to get only shadows in a render: Invisible shader of course doesn't work because it's a Shader Lab shader... turning on transparency for the object eliminates the shadow.
- Difference Matte: This is a trick where basically you render everything out - characters/props + ground plane - and then render a second pass with only the ground plane. You then take the "difference" of these two passes and it yields the objects and shadows. That then becomes the mask for the original foreground layer and gets composited over the live footage. Sounds cool, right? Well, it becomes difficult to get a clean comp because there are still adjustments that have to be made to the difference matte as far as shadow transparency goes, and it can often affect the 3D elements.
Does anyone know of a way to achieve this in Dribble? |
Post by jayr // Dec 26, 2007, 11:05am
|
jayr
Total Posts: 1074
|
i haven't done much composting so if i'm missing smething obvious please tell me, but....
Couldn't you try shading the objects with a matt green shader then 'green screen' them out? |
Post by frank // Dec 26, 2007, 11:21am
|
frank
Total Posts: 709
|
Jayr - that is actually a good idea! The only issue with it is that it doesn't render the shadow that would be underneath the object - so whenever I do any blurring, it can give a "halo" type thing around the 3D object, because after keying it out there is no pixel data directly beneath the object, and that gets "spread" out past the object above.
HRMMM..... but now that I think about it, I may not need to do much blurring. The only reason I did it before was to get a touch more realism from the standard tS-rendered shadows, and I likely won't need to do that with the Dribble-rendered versions. Didn't think about that. :)
Certainly worth a try! Thanks, Jayr! |
Post by frank // Dec 26, 2007, 5:01pm
|
frank
Total Posts: 709
|
Gave that a try and it didn't work too well. The subtle soft shadows that Dribble generates really don't show up too smoothly once the green is keyed out - they just look harsh.
As much as I don't like compositing the footage in-renderer, ie. 3D elements rendered with the live footage set as the background, (after all, that's what apps like After Effects are for), if we had a basic shadow-catcher shader, that would work, so long as the lighting is matched well (HDRI / image-based lighting would do the trick).
A really cool solution would be a way to alpha-map the shadows. ie. if Dribble could render the shadows and set the alpha channel values for each shadow pixel, with the value/transparency depending on how light or dark that part of the shadow is. |
Post by Emmanuel // Dec 26, 2007, 11:46pm
|
Emmanuel
Total Posts: 439
|
That's a very good suggestion for future Dribble features.
Good idea Frank ! |
Post by SiW // Dec 29, 2007, 7:13pm
|
SiW
Total Posts: 298
|
There's a *lot* of flexibility in Renderman - and 3Delight specifically - for compositing, and it's something I'm looking at myself. For example, we should be able to render out the shadows out to separate channels if we want. The trick is being able to expose all this stuff in a way that is easy to work with.
I'm not opposed to some sort of "expert mode" for new or very niche features as long as people understood that there'd be no handholding. |
Post by frank // Dec 29, 2007, 7:37pm
|
frank
Total Posts: 709
|
I'm not opposed to some sort of "expert mode" for new or very niche features as long as people understood that there'd be no handholding.
Hey hey! Sounds good to me! |
|
