Digital matte - applying painting to geometry

[dstipes]

Hello all,

I am familiar with the older traditional (analog) ways of doing matte paintings.

Can anyone provide more explanation of the newer 3D digital techniques for matte painting compositing?

Building 3D geometry for the matte solves some perspective and placement issues.

However, I am confused as to the application of the painting to the geometry.

Are painters using front projection to put one large digital painting onto all the shapes at once? Or, are the painted images being applied to each surface like typical texture maps? A combination of these?

Any clarification would be appreciated.

Can anyone post some visual examples of the steps?

Thank you.

David

[batera33]

Hi Dstipes,..

Well, I`m not an expert about 3D, but I think you should see this link
http://67.15.36.49/ffa/tutorials/tutorialsoverviews.asp

There is many ways to manipulate the textures maps, in this link you can see some...

But, in my opinion, they work with many maps, applied to the geometry...[/url]

[AliK]

You can obviously use your digital matte painting as background in any program.
You can also use it as a plane in compositing or 3d programs.

To take this further you can also model the scene and project the painting over the 3d geometries.
This will give you some freedom for virtual camera motion.

For more freedom of camera movement, you might also paint some texture for some part of the geometry.

The maximum freedom will be achieve by doing this come and forth between painting and the rendering results. In another word you paint and then project over 3d, then render the 3d scene and then paint it again…

Take another look at www.matteworld.com (http://www.matteworld.com) , www.tchook.fr.st (http://www.tchook.fr.st)
and the link batera33 submited.

[rrische]

Hey David,
A good explanation....picture a stage with a bunch of plain cardboard boxes laid out all over the floor, and you want 'em to look like buildings. You have a motion control camera all set up, and that camera is actually
going to shoot the shot. Now you light the scene to make sure you have
shadows and highlights in the correct place.
Next, you set up "invisible" still cameras to look at the boxes closer up, making sure that each portion of the box that is seen by the main camera is covered in the angles being seen by the invisible cameras. Then you take
a series of snapshots of the boxes from the invisible cameras.

Next, you bring these snapshots into Photoshop, and paint and season to taste. They should (hopefully!) look like real buildings when you finish, with
the same lighting direction that you established when you lit on stage.

Now, the fun part....you load your snapshots back into your invisible
cameras, and turn the cameras into slide projectors! Now they're projecting your gorgeous building artwork onto the flat boring cardboard
sides, exactly in the same orientation as your original snapshots.
Then you turn off all the lights on your stage and go back to your motion
control camera, and get ready to shoot. The last stage is to snap your
fingers and make the cardboard boxes invisible. Then you shoot the shot.

That's pretty much it. Different 3D programs have different names for
all this, but it's pretty much the same. Where you place your invisible
cameras is completely dependent on how extreme the camera move
is. If the move is too extreme, your nifty projections will tear and
stretch and begin to slide off. If this happens, you have to go back
and adjust the problematic invisible camera position and adjust the
painting too.

On "Day After Tomorrow" we had one shot with more that 50 different
separate camera projections in it. It's during the superfreeze in New
York, where you see the frost forming and running down the side of
the Empire State Building. The camera is racing down the side of the
building trying to keep pace with the ice as it forms. The camera travels
a LONG distance, it also tilts up, pans, and pulls out during the move.
To cover the New York background for the entire length of the move,
we had to split the BG geometry into many many pieces. The Empire
State Building itself had six different camera projections on it for that
one shot.
You can do very complex shots with this approach, but the technique
itself is really pretty simple. And cool!

[rrische]

Here's one of our many camera projection matte shots for "Day After".

http://www.mattepainting.org/plugins/p17_image_gallery/images/454.jpg

[Tchook]

Thx for sharing rrisch,
im a very big fan of this movie (for the sfx quality), the tornado sequence
in LA really knock me out.
i m just curious , how do you manage to render this huge amount of geometry?
i mean do you guys separate each building in seperates scenes than inport them as "reference" in a master scene ,which is separated in layer for render?
or ?
thx in advance

[rrische]

This production utilized a 3D scan of the island of Manhattan (yep!
The whole damn city was scanned using Lidar technology.) You can
see a picture of the model here...http://www.visuraimaging.com/day11.html

This picture shows the model dressed with snow,
but it was bare when we got it. All the snow and building detail was
added in the matte painting.
As you can see, this is a ridiculously HUGE model! What we did was decide
on a shot-by-shot basis, what portions we would need for a given shot. We
deleted any geometry that wouldn't be seen by the camera.
If a shot used, say, 14 different projection cameras, we would make 14
different copies of the 3D project, and delete any geometry that wasn't seen by that particular camera. This started to make the job somewhat manageable. Then each separate render would be composited with all the others (and fog passes etc.) in After Effects.
Still, these shots REALLY were heavy. There was no way around it.

[rrische]

If you want to learn more about Lidar Imaging, click here...

http://www.bluesky-world.com/lidar.html

[Tchook]

thx for the links..
wow scaning the whole city, i now know what you mean by ridiculously huge.
the displacement map are incredibly accurate.
does this scan where accurate or is there a painfull process of model cleaning?


if i m too pushy tell me, i will shut my mouth right away.... :wink:

[thenextside]

Rick,
That is probably the best explanation I have heard so far. Thank you. Just a question about the invisible camera snap shots. Do they need to be flat on shots (meaning, with no perspective)? And I am assuming that you need to do a projection for each individual box? Have I understood correctly?
Thanks so much for sharing you expertise.
Cheers,
Tim

[rrische]

The simplest solution is always the best. I always want to start
with the easiest approach, but switch to something more complex
when it's necessary.
With camera mapping, the simplest approach is to use ONE single
map taken from the actual shot camera- one camera, one painting.
This map will look just like the shot, with perspective, lighting etc.
Depending on your camera move, this may or may not be enough.
If you run into technical problems (stretching, double mapping etc.)
then you need to start splitting the scene into smaller sections and
projecting those separately, and combining them together in the composite.
If you're STILL having problems getting a good result, then it's time to
start moving your projections off-axis from the shot camera. This can
get a bit tricky for the artist, as your paintings don't quite look like the
shot anymore. You need to do a bit more back and forth testing to
make sure that your paintings blend together and look correct from
the shot camera.
Probably the most extreme approach (and the least desirable
approach from the standpoint of matte painting) is to set up "front on"
or orthographic views of the geometry. You would do this if your camera
move is SO extreme that no perspective view would hold up (for example,
a 360 degree rotation around the object). I've only had to do this once.
It worked out, but it's difficult. A lot of testing is necessary to assure
you get a good result.

Hope this helps.

Oh, and the model of New York didn't need any cleanup per se, but it
DID need a lot of technical tweaking. It had flipped normals, whole
sections of the city wouldn't accept projections or textures. This had
to be fixed before we could do shots with it.

[thenextside]

This is really helpful. Is the camera move happening in the 3D program or the compositor? When you bring it into the compositor, is it a matter of sequencing the rendered out segments together? So when one sequence starts to fall apart the next one picks up and so on. Thanks so much for taking the time to explain. I have been trying to find these answers for some time.
Cheers
Tim

[Novalism]

very usefull information and insights
tnx for sharing indeed

i wish there where some more complex tutorials arround. As long its a basic box i can map... if i go further than basic geometry .. bah
how do you deside what to model and what not.
:roll:

[alasdair-m]

There is Janine's tutorial in the tutorials page on this site which is very good, but I agree with you I am trying to find a good maya tutorial for camera mapping, so if anyone knows of one or if someone would like to take the time out to make one I, and many many others I'm sure, would be eternally grateful. Cheers :D

[tobiasth]

take a look at this site:
http://www.mikebreymann.com/
> tools > tutorials > camera mapping

or this one:
http://www.a3d.pwp.blueyonder.co.uk/maya/camera_projection.html

hope it helps 8)

[alasdair-m]

Thank you very much for that, there goes my weekend. :D

[kaz]

what is the older analog ? does this mean just the original of hand painting cause i thought analog was a term in mechanical or tape wise

[rrische]

analog=not digital

[kaz]

what software does camera mapping?

[rrische]

3D Studio Max
Lightwave 3D
ElectricImage
Maya
Cinema 4D

These are the ones that come to mind.

[Nazgul]

Great stuff!!!

I love to see the map you painted for the projections..

[Nazgul]

The snow in "The Day after Tommorow" was all 2d matte painting projected on 3d geometry. Are there any CG generated snow or particle effects?