I played around with this project a few months ago, and it was cool, but with some limitations. The microphone I had to use didn't work from very far away (that was my fault for sourcing a crap microphone), and the voice recognition was finicky (even after getting close enough to the mic). I know the voice recognition is just another piece of software that they integrated, but since the project relies on it so much I think it would be well served if there was an easy way to train the voice model to understand me.

Simple plugins are easy to write for things like checking sports scores, or anything with an easily accessible API, and this is a huge plus.

You can swap out the voice recognition portion. The default is PocketSphinx but there's also a Google Speech to Text implementation.

https://github.com/jasperproject/jasper-client/blob/master/c...

I wanted to install this on my existing raspbmc install but the instructions are fairly lengthy - lots of steps and dependencies - and I wasn't sure if the pi is actually fast enough to run this and control XBMC at the same time, which is what I really wanted to do. It does look fantastic though.

It just occurred to me that a pi is so cheap I could get one and install the SD card version of Jasper and control XBMC over its API :)

Isn't "Jasper" the pretty well-established name for a Java report library? That's certainly the first thing that comes to mind for me.

Whereas the first thing for me is the Jasper prover (http://www.jasper-da.com/)

I don't think it's really a problem. If you're worried, just don't build them into the same product :-)

This is interesting, and I certainly hope to see updates in the future. For me, for now however, I found wit.ai to be so simple to get up and running that I can't see myself moving to something like this - even though I can see significant advantages (local data only, for one).

Does anybody know if it's possible to have it pick up "arbitrary" words instead of predefined commands? E.g. have it transcribe a conversation? The detection wouldn't have to be super reliable for what I'm thinking of.

A repost of https://news.ycombinator.com/item?id=7546858.

Curious why they decided to create a separate GitHub user for the project instead of using an organization.

I'm not sure why they did that, but I really wish I had multiple github accounts rather than one.

When integrating with external services (like a CI server) things will request read or write access to "public" or "private" repos. There's no distinction between my repos and my organisations repos. I might be happy allowing access to my personal ones, but I cannot just give access to my companies repos. Even if I wanted to for single repos, I can't justify allowing access to them all. The permissions model of github is far too coarse.

Agreed with those complaints. However, as a workaround, I find a lot of the time you can just add the web hook manually for a service to a repo manually and things work just fine.

Cool project and everything, but considering my name, it's extremely freaky to see text like "Click on the guides below to learn how to build your own Jasper." I'm too distracted by it to understand what else is going on.

Whoa! Jasper posts to HN!

Ha! And people laughed at me for believing singularity would happen this year!