The first project (your "this" link) is mine. To clarify, the sound in that demo is generated on the computer not on board the device. The device acts as a USB MIDI controller. There is a standard for USB MIDI so you can just plug it into any computer with suitable software and it should work - but there is no on board sound. This is also how the Moolodeon works IIRC but he used a different chipset.
Having said that I'm still intending to try to implement on board sound. Having done some research I think it can be done on the same chip that I used for that project, which is nice because those chips are fairly cheap and can also do the USB thing. I would have to reduce the sound quality a bit for on board generation to achieve a high enough polyphony. The main sticking point though is time: I have two small children and lots of other things I want to get done (like practising the melodeon!).
However if it worked and could be fitted into a real bellows instrument with a pressure sensor (like the Moolodeon) I think it would be fun to play.
The problem with any project like this is the amount of work required to take it from a hobby project to something that you would actually be able to use regularly and then another large amount of work to make it into something that somebody else could use easily. It was only a few hours work for me to write the code for that demo but if I want to add a new layout (for example) I have to modify some numbers in the code, recompile and re-flash the chip. Polishing it up into something with the ease of use of the Streb or the Roland thing is a lot of work.