As was discussed in one of the sessions - having a local www.techshop.ws with access to CNC machines, giant band saws, laser cutters, arc welders etc would be AWESOME. For those of us without any big machines or even a fretsaw here is how to cut a big sheet of perspex with a coping saw. Relay cutting allows you to cut a channel big enough for the 'neck' of the coping saw to cut down.. hence you can cut as deep as you need. Initially I tried to make a sheet of transparent silicone by putting sealant onto mono (baking) paper. Then spreading it with a squeegee. Unfortunately this particular sealant was too effective, and would not seperate from the paper - rendering it decidedly opaque. I have successfully used this technique in the past with other silicone sealants.. but for this purpose it was a bust. Next it's time to do some cooking (boiling water). The mixture is half a cup of water and one teaspoon of gelatin. Along with roughly 1/2 cubic cm of phthalo green acyrlic paint (Phthalocyanine). The mixture should end up pretty liquid - the aim is to create a colluloid gel rather than a solid jelly. Mix this thoroughly and pour carefully into the mould. As we have no silicone sheet, plastic wrap will have to do the job. The main problem with this material is it is a little pliant and tends to stretch when you 'drag' your finger on the surface. The finger contacts are much better with the gel. With pure liquids the plastic tends to deform in a much wider region around the finger - however the lighting was not ideal in the test setup, making the shadows of the hand a little ambiguous. I ended up making a gel sheet to the limits of the perspex I had cut. Here is a video of it working with the blob tracking, and my multi-touch framework. It was a bit spotty, not even close to the NextWindow hardware but I used a thicker plastic than the plastic wrap above which helped, and slightly more water which stopped the tearing (e.g. the gel was more liquid). Here is a video, it's in two parts. The first half is the processed webcam video, you can the see the shadow of my hands and a contact as a white spot under one of my fingers. Because the gel tints the hand when it is above there is no false tracking on the hand that is moving. The second half of the video is me using this input on the multi-touch framework. As a single point click it was fine, but this was the best footage of dragging. This tended to stretch the plastic too much. Where the gel is 'torn' it leaves a gap which results in false skin tracking where your hand passes over the gap (hence the accidental drag just before the resize). I think the best plan eventually is to use a capacitive sheet on the desktop for touch, and the webcam above purely to generate the hand shadows onscreen.