The first-generation iPad is optimized for viewing 3D objects with fewer than 20,000 moderately textured polygons. Second-generation and later iPads and OS X Mavericks support viewing 3D objects in a completed book with up to 50,000 moderately textured polygons. More complicated objects may result in an increased load time, slower behavior, or a blurry image.
Use simple lighting with ambient, omni, and directional lights. For best performance, limit the scene to three or four lights.
Any textures you use that are greater than 1024 x 1024 pixels will be resized to 1024 x 1024.
Be sure to add a camera to your object. This will be used for the default view in your iBooks Author book.
A camera can use a perspective projection or an orthographic projection. These are ways of projecting a 3D scene onto a 2D surface.
In a perspective projection, objects close to the camera appear larger, and objects far away from the camera appear smaller. In an orthographic projection, objects appear the same size regardless of how far they are from the camera.
It's best to use a perspective projection in COLLADA (.dae) files you add to iBooks Author, as it mimics the eye's natural perception of the world. Orthographic projections are rarely used.
For more information, see the COLLADA Digital Asset Schema Specification.
Some modeling programs will get better results using the OpenCOLLADA exporters in place of their native exporters. These exporters are available on the OpenCOLLADA download page.
To confirm your 3D object behaves as expected, be sure to preview your book on an iPad or in OS X Mavericks before publishing or exporting your book.
For more information on the COLLADA format, see the COLLADA Digital Asset & FX Exchange Schema.