It's been a while since I've had a little project with one of the Nao robots at the museum, but here's a quick one that I was asked to create on short notice. Over the past 4 or 5 years I've put together performances for big crowds, trade show booths, and private greetings, so I've gotten pretty good at programming some quick motions, adding some voice and sound effects, and then refining until its reasonably smooth.
Honestly, one of the biggest challenges with this platform is getting the voice just right, since I'm working with a text to speech engine. Sometimes this means tweaking the speed or using homonyms with nicer pronunciations ("humans" always sounds better as "hugh, mens".) Occasionally though it even means getting into adjustments to the pitch, emphasis, and pauses.
One of the most valuable tools I've found to help with refining both the vocals and the movements is to record video so I can really dissect things (in fact, most of the videos on this blog are saved from testing.) It can be incredibly useful to rewatch little snippets over and over to get a feel for what's good or bad as opposed to watching it in real time and trying to remember the details after the fact. It also helps to compare versions as you make small adjustments so that you can decide which option is better.
