Using AVISynth to get a tiled Multi-Viewer 3D video - Part 1

Many people ask me if they need to buy special software in order to create 3D videos for glasses-free displays. Well, the first thing you need is some kind of 3D application: 3DS Max and Cinema 4D are not for free I know, still you could do something using blender for example.

As far as i know, there is no plug-in for this tool until now, but it should not be difficult to write one.
On this post however, I will be showing how to get the videos together. It means, you have already rendered your views using blender or other software.

If you ask yourself which resolution should I use? Well Tridelity has created a very useful document on Multi-Viewer 3D resolutions which answers this question.
For this example, you will need:

• VirtualDub
• AVISynth
• A video codec, I recommend ffdshow or xvid
• 5 picture sequences or 5 videos (each video is one perspective)

In the first place, we should read the picture sequences and name them on the script. Create a text file and name it for example ShowReel.avs
Open this file on a text editor, I recommend using Scite.
In the following example, I have pictures in full HD resolution. The images were rendered from 0 to 1200 but I want just to use the images 970 till 1042.

a5 = ImageSource("view01_%04d.jpg", start=970, end=1042, fps=25)
a4 = ImageSource("view02_%04d.jpg", start=970, end=1042, fps=25)
a3 = ImageSource("view03_%04d.jpg", start=970, end=1042, fps=25)
a2 = ImageSource("view04_%04d.jpg", start=970, end=1042, fps=25)
a1 = ImageSource("view05_%04d.jpg", start=970, end=1042, fps=25)

The name of the images are viewXX_YYYY.jpg where XX is the view number, 01 being the right camera view and 05 being the last camera from right to left. YYYY is just the frame counter which I mask using %04d (0 = leading zeros, 4 = 4 digits, d = decimal).

I use the AVISynth command to read images called ImageSource, here you can define the starting and ending image numbers, as well as the frame rate using the fps parameter.
Notice that if the images are in Full HD, we will need to resize them. According to Tridelity, we should get only half of the resolution in order to make sure we don’t need expensive hardware to play the videos.

We can easily do it by adding a resize command:

a5 = ImageSource("view01_%04d.jpg", start=970, end=1042, fps=25).LanczosResize(960, 540)

This will do the job and resize the images to half of the size.

On the next post, we will continue from here. If you however want to have a quick look on the video, please use this script in order to get a side-by-side (half width) 3D video:

a5 = ImageSource("view01_%04d.jpg", start=970, end=1042, fps=25) ).LanczosResize(960, 1080)
a4 = ImageSource("view02_%04d.jpg", start=970, end=1042, fps=25) ).LanczosResize(960, 1080)
StackHorizontal(a5, a4)

Now just load it on VirtualDub, and voila! You can now save your first stereo video which can be uploaded to YouTube for example.

See you on the next post.