This is a project describing a workflow to create 3D scenes from 2D images.
When deciding how to paint the individual layers here, try and think of breaking down the objects in the image into pieces you might cut out of paper to build a 3D object. I like to think of building a house out of playing cards. The cartoon South Park is the best example of what we're going for here.
- Start by creating a new image, 900 x 600 pixels (just to keep it simple, I'm using the "Photo" preset). Make sure Color Mode is 16-bit and Background Contents is set to Transparent.
- Create different layers to represent a ground plane and a couple of objects.
- Open your
~/Documents/maya/scripts/folder from a terminal window. - Type
git clone https://github.com/zklinger2000/maya-python-imageCard.gitto download the scripts.
- Open up Maya to a blank scene.
- Load the python script imageCardsImport.py into the script editor.
- Make sure all the folder paths in gravModTools.py and imageCardsImport.py are correctly assigned to your local folders.
- Highlight the entire script in the script editor and hit
Enter. (There are multiple ways to run scripts in Maya other than this one) - Depending on how many images you have in the folder and how big they are, this may take a little time.
- Once the script has completed running, if there are no errors (usually having to do with improper folder name escaping, etc.), you will see a big black, yellow, red and green blob filling up your viewport.
- Zoom out and click on the masterLayer in the Render layer tab to see the imageCards.
- TIP: In the viewport panel, go to Renderer -> Viewport 2.0 -> [] and change Transparency Algorithm to Depth Peeling.
- Only add lights to the master/beauty render layer!
- Use whatever lights you desire, but simple spot, point and directional lights save tons of rendering time.
- Make sure you turn "default light" off in the Render Globals.
- Use mentalray as your renderer for each pass except Z-Depth.
- Final Gather is optional.
- Much more detail is required here
- Add a simple camera to your scene.
- Turn off all the other render layers and render out the color/beauty pass.
Beauty Pass
Ambient Occlusion Pass
Z-Depth Pass
Position Pass
- Use Software rendering for Z-Depth pass
- Tons of detail here...
Final Comp
There are still some issues with premultiplying around the edges, but that's due to the images being drawn in Photoshop without proper alpha channels. I've found that just simply adding a lot of detail and drawing with a thick outline, comic book style, they get hidden mostly. Otherwise, it's been an interesting exercise. Thanks for reading! I'd love to see how people can make this better.
