I usually complain about apps being available to one platform instead of being webapps or at least having some API to avoid cutting out all the other users, but at least this time I see the reasoning behind it. You take a photo, it tells you what it sees. Nice, how does it do it? any chance of seeing it on other platforms?

Thanks! Frank is an ongoing experiment (some more info on that here http://blog.terrarium.io). We are currently gauging demand, so if enough people enjoy him he may be developed further.

Interesting stuff. Even though it takes a while for the caption to be generated, the humor makes up for it (most of the time).

I'm assuming you guys have plans to integrate this into your other product (Peddl)?

So what's the technology behind it ? How is the caption being generated, I presume some sort of mechanical turk / Franks ?

I had the same idea, captions are human generated via the turk.