Green Screen Directly in TG

Started by TheBadger, July 13, 2014, 05:18:53 AM

Previous topic - Next topic

TheBadger

Hi,

K, before I ask my question, here is the reason I'm asking...

If you guys remember Moodflow's work on using photos to add tons of info to renders, as seen here : http://www.moodflow.com/wordpress_moodflow/wp-content/uploads/2010/12/7a237eac9ee886e94679e8f7e06337481.jpg and http://www.moodflow.com/wordpress_moodflow/wp-content/uploads/2010/12/monolith-fields.jpg and lastly: http://www.moodflow.com/wordpress_moodflow/wp-content/uploads/2010/12/2a540df1035625befa620691b82d4379.jpg
Then you will also remember that TG will take color info from the image and blend it in the render.

So rather than comping and color correcting, that is all in a good place right out of the render-er.

My questions are

1) Just as you use a photo in moodflow's method, why couldn't I use a video/image sequence?

2) Rather than comping a render to a matt, couldn't I just project a green screen video onto a plane or card object and, and somehow cause the part that is green to be transparent. Then make my camera line up in the proper place and angle, and then render through the projection (so to speak)?

To be clear, What I want to do is project a video/image sequence of foreground activity onto a card or plane or some TG object in the shape of the recordings aspect ratio. Next I want to make the part of the video that is green screen transparent, so that I can see through to the background in TG.

Next I want to render a scene, with the part I would have comped in later already in my scene JUST AS in the op links above, but with video rather than a still.

3) Why not?

Thank you if you understand what I'm getting at here.

Cheers.

It has been eaten.

Oshyan

I'm pretty sure the 1st and 3rd images are just photos or, if they're some kind of composite, probably more like Photoshop. I don't see any likely TG elements in them, unfortunately. His gallery *is* clearly a mix of different kinds of art/media, including TG and photographs. The 2nd (middle) image is indeed TG though, the foreground just uses a texture map from a photograph.

Now as far as what you want to do in TG, it would probably be much better to just create your alpha in another app and import a sequence with existing alpha in TG. You *could* use the Create Transparency function on the Effects tab of an Image Map Shader and key it to green, but the control you get is probably not as good as you'd get in a dedicated image editor or compositing app.

Other than that what you're talking about sounds feasible. I'm not certain TG supports image maps for animation, but I think it does. You know of course that it supports object sequences, so it'd be the same approach, use a wildcard in the file name, e.g. texture_%04d.tif

But that isn't really what Moodflow did. He used photos as textures on the terrain, much as many others have done, to get better variety or more realistic colors in surface textures. This could also be done for skies, if applied to the Background Sphere, or a card object (if you wanted it in front of the Atmosphere, say).

- Oshyan

TheBadger

Hi Oshyan

Here is the test I did a long way back. http://www.planetside.co.uk/forums/index.php/topic,13385.msg132061.html#msg132061
I think I'll start with the same thing but with a video, and try to get it in the way mentioned above. Im going to try and find a good stock and hopefully it will already be keyed, or at least lend its self well to being keyed. Then Ill try the various ways this could be done. I'll start. with the way in my own image link, and move on from there.

Going to try and find some stock with a nice breeze happening. Would like to find some grass like in the first image link in the OP by mood flow.

About those images in the OP. You may be right about part of it. But I think he did use TG for some part of the first image. I had thought that the first image was all TG using this method. But...
It has been eaten.

TheBadger

OK, so doing an image sequence like moodflows still method should be pretty straight forward.

but now I'm a little confused on the other possibilities when a camera move is added. Don't know if it will look 'good', just want to see if I can make some things work.

1) when you project an image, TG uses that images colors. I can't really explain how. But when projecting an image it does not need to be color corrected to mach the rest of the scene, it just sorta blends in. However, would that be the case with a card object somehow?.. If that makes sense? In moodflows method, the projection is to the terrain. But the card object is detached from the terrain. So color-wise, is this going to be different than projecting a 2d image to 3D terrain?

2) I understand the mechanics of all the options. I mean I can see in my mind how to put this together. But I get pretty confused when I start thinking about camera moves...

If I project an image onto a card object (such as one of the possible methods listed above earlier) Can I, a.) somehow "lock" the card to the camera so that it is always in the same place in front of the camera, no mater where I point the camera. And b.) If the projection to the card object is a video with a camera move, and I have a camera tracking .chan for my TG camera (where I am rendering through the card object (as Oshyan proposed above)... I can get that to look correct right in terms of how the TG background moves relative to the camera, relative to how the video projection on the card looks?

This is now very confusing ;D

Obviously, the moodflow way won't allow me to animate the cameras postion, except perhaps for a little "on the shoulder" camera shake. But the card object offers more opportunity when used with camera tracking I believe.

Now to be clear, really all I am talking about doing is what others use TG spherical renders for in Maya and other packages (also compositing). That is, they render out a scene as a spherical render, and then use that in maya to light their characters and provide backgrounds. And in the compositing sense, the card object thing, would be a matt in front. But I really want to see if I can do this in TG and what if any benefit there would be.

Just experimenting. Did not see any other posts using video since getting TG2, so I thought why not?

@Upon Infinity
If you are reading this, I think I need to review the information in those threads you started where we talked all about how you used your still images on a card object, or imported cube I think it was. Can you post a link(s) here? I think I can do what Im trying to do, but I need to re-read a bunch of stuff. Im just trying to use video instead of stills in a number of ways. But I don't want to have to re experiment on parts that are already documented in the forums.

@fleetwood
http://www.planetside.co.uk/forums/index.php/topic,17671.0.html?PHPSESSID=3e3be323ee0b49c9b49465f609d0b28f
Any thoughts about doing this with video?

Thanks for playing along guys!
It has been eaten.

TheBadger

#4
Hey, this will help with showing some of the ideas Im playing with http://www.youtube.com/watch?v=GmnqcgonFIE

And this kinda shows https://one1more2time3.files.wordpress.com/2010/06/bambi-bb45.jpg?w=510&h=229

Finally found video that shows one of the classic processes I want to play with!! http://www.youtube.com/watch?v=dP3ElTbrSBc start at 11:18 - 11:53.

Look at this http://www.youtube.com/watch?v=YdHTlUGN1zw !
http://3.bp.blogspot.com/-oAKVIAFV6S4/Ufq-M6V3fhI/AAAAAAAABYw/xFPguGWEryE/s400/cottage-snowwhite1.jpg
It has been eaten.

bigben

The problem with camera movement and images projected onto cards is that you have to be careful to keep the card relatively perpendicular to the camera otherwise you will get some weird perspective shifts.  Projecting onto the terrain may limit your camera movement, but regardless of what method you use there will be limitations regarding where you can move/point your render camera.  XREZ always have lots of clever stuff. It should be possible to do this: https://vimeo.com/8630144 or some of this: http://www.xrez.com/cw_video_pano/ in TG.  Higres images/image sequences projected onto DEMs/laser scans to allow for some virtual dolly effects.

TheBadger

#6
Ben

:o :o :o :o :o :o :o :o :o :o :o :o :o :o TELL US MORE!!

That is crazy! I want to know more!

Link one is state of the art! I never saw a maya render like that before!
link two "Full Dome theatrical projection" I nutted. Im sorry, but I did. Its a mess  :o
It has been eaten.

bigben

Link1 is so 2010  ;)  These guys do lots of stuff with high res imagery (usually image stitching). There's money in movies so they can afford the right gear (or to make their own)  You'll notice that the camera moves forward and down so that you don't see that there's nothing projected behind the ridges, so it's basically a video projected onto a DEM through a camera at the same virtual location.  If you project the image through the render camera it shouldn't be too hard to line things up, then just leave the camera in place and create a new render camera for the motion.