I'm not a part of the dev team, so I couldn't say for sure, but the OpenNI system seems to returns a 16bit depth value for each pixel, so this is likely somehow encoded into a 24bit RGB stream. If you take a quick recording then sample the colors, for me the R was always 0, the G always 3 or 7 and the B varied widely. That makes me think that the depth value is encoded into the green and blue channels. Possibly MSB in G and LSB in B. You could do some testing yourself with objects at known distances and see if you can find a correlation between the values encoded in the video and the actual distances.

Just out of curiosity, what software were you planning on capturing depth information in before converting it to the format iPi Studio expects?