Exporting Models from Max into Autostereoscopic Animations

To get models created in 3ds Max to export into a format that can be imported into an autostereoscopic 3d (A3d) compositing program, I use an A3D camera plug-in, which you can download from 3D International Europe GmbH, formerly Visumotion.

This special camera is actually 9 cameras side-by-side (you could actually make one if you have max ninja skills). [3D Artist, #19 has an artical on how to create a lenticular 3D camera setup in 3ds max] Each camera sees the frame from a slightly different horizontal perspective - but all focused on the same point.

You can then export the scene by saving it a a TGA image sequence. What you are doing is saving 9 separate images per one frame of video. So, for a video with 30 fames per second (fps), multiplied by 9 full-scale images per frame, you have 270 frames per one second of autostereoscopic video.

When you want to combine the nine separate TGA image-sequences back into a playable format, the next step is to compress all the video streams using H.264 and wrap them in an MP4.

Another Player-like software package will read all nine perspective frames per frame and rasterize them together according to the rule of the lenticular overlay. The image it creates, when seen through the special lenticular overlay, appears as 3d to human viewers.

I think this technology is amazing. Sure, there are still some issues with the technology (one being that you need to be viewing the monitor in certain spots, and if you move your head left or right while watching the 3d monitor, the image will go out of focus. I explain the 3d viewing planes in another post).

Brief 9-View Multiview Autosterescopic Video Frames Summary

How does it work? is a question I hear often. Here's a brief answer: Normal 2D video is usually 30 frames per second. That means for each second of video, you actually see 30 still pictures in a row, each displayed for 1/30th of a second.

For an autostereoscopic video, each frame consists of 9 different pictures, each picture from a slightly different viewing angle. So, for each second of an autostereoscopic video, you have 270 pictures.

For each frame, each of the 9 pictures is rasterized into a single image according to the rule of the 3D lenticular overlay. When that rasterized image is displayed by the LCD monitor, through the lenticular overlay, it then appears as an autostereoscopic 3D image.