KEIICHI MATSUDA

password = unit15

Test for anaglyph 3D using rig below. It can be viewed using red/blue anaglyph glasses; it actually looks pretty good.

The most successful bits are the long view over the rooftops, the pitched roof towards the end, and the strange creature on the floor. With a stereo base of 80mm, objects between 3 and 10m away seem to give the best 3D impression.

There are also some things that dont work so well:

• Using anaglyph 3D causes white areas to ‘flash’ uncomfortably, and also causes problems with red/blue objects.
• It is impossible to get both cameras to record exactly the same frames. Assuming the worst case scenario when recording at 25fps, the cameras may be 1/50th of a second off, which makes a noticeable difference. Nothing that can be done about this as far as I can work out (unless I use frame blending on one of the clips, which seems a bit over the top).
• small near objects against far objects feel uncomfortable. While the exhaust upright at 00:20 looks fine, the washing line looks horrible.
• objects tilted towards the camera (roof at the end) look better than layered planes perpendicular to the camera (roofscape).
• It looks MUCH better in HD.

AE’s 3D Glasses effect makes merging the two videos really easy. It also allows you to set and keyframe ‘convergence’ which shifts the horizontal position of the videos, thereby changing which position on the z-axis corresponds to the plane of the screen. This also means you can ‘focus’ the 3D on objects. I did this for the hands, but it obviously disrupts the background.

Most of the tweaking is to make the video comfortable for the eyes, and to make the 3D as deep and detailed as possible. I’m pretty happy with the stereo base for most of the shots I’m planning, but it may cause problems for near-field objects (such as hands). Combining convergence manipulation and depth of field may solve this problem, but will be hard to integrate with a point-of-view shot.A 3rd person perspective may be better…

Although its really interesting to have a go using 3D, it may eventually become too time-consuming. I will have to sacrifice some colour, and will not be able to pre-render anything. Shooting in 3D also means that I will have to ROTOSCOPE EVERYTHING TWICE!! It should be easy for far-away objects (a copy/paste job), but harder for near-field (which are the ones which need to be rotoscoped anyway).

Im shooting on Sunday, but undecided on whether to produce in 3D or not. Since I have the chance, I will shoot in 3D anyway, but may end up using only one of the feeds. I dont need much footage, so will shoot both first and 3rd person perspectives, then see what looks best in the edit (and in 3D). Maybe a normal version for the submission and 3D for the exhibition?

In other news, I wasn’t able to get permission to film in Liverpool street station due to ‘security procedure’. Now trying to think of another location for the final meeting and panoramic end shot. Any ideas?