JiBotVoice


Joi's Wiki has moved to [WWW]a new SocialText wiki
This Wiki is now disabled.

This is a placeholder for the future JiBot Voice module, as imagined by termie and bunnywabbit (with creative input by adamhill and yonderboy)
10:28 < termie> yonderboy: yup, and... uh... VoIP
10:28 < patfm> +have fun suw
10:28 < yonderboy> +termie, j/k
10:28 < patfm> +termie: whoa, for real?
10:28 < patfm> +jibot speaks?
10:28 < Suw> -i shall pat :D
10:28 < termie> patfm: surewhynot ;)
10:28 < patfm> +heh
10:28 < patfm> +make sure he speak ainu
10:28 < bunny_home> -termie: what tts are you going to use?
10:28 < bunny_home> -:-)
10:28 < yonderboy> +termie, rss urls as user metadata would be nice though
10:29 < termie> bunny_home: the good one
10:29 < bunny_home> -good... coz the one I heard on osx clearly sucked
10:29 < termie> bunny_home: yeah, for sure
10:29 < bunny_home> -I much prefer scansoft tts (at least, the one that comes with dragon)
10:30 < bunny_home> -I think scansoft were top for tts a couple of years back in a review I saw
10:30  * termie is a Yes Man
10:30 < bunny_home> -and nuance top for stt
10:30 < termie> bunny_home: I would definitely agree
10:30 < yonderboy> +crysflame, still awake?
10:30 < termie> bunny_home: (but I haven't tried any of those)
10:31 < bunny_home> -but nuance don't do any "desktop user" speech recognition thingies, so joe lambda is stuck with scansoft
10:31 < bunny_home> -(which is, honestly, really good enough)
10:31 < termie> bunny_home: (and this was a totally hypothetical situation that won't happen for... a long time)
10:31 < bunny_home> -(at least dragon on pc, I'll let you know when I've tried viavoice on mac)
10:31 < bunny_home> -termie: what was a hypothetical situation?
10:31 < termie> bunny_home: alright, you can be the official stt / tts QA for jibot
10:31 < bunny_home> -?
10:32 < termie> bunny_home: i said I was adding VoIP to jibot, as a joke
10:32 < bunny_home> -oh.
10:32 < yonderboy> +heh
10:32 < bunny_home> -:-|
10:32 < bunny_home> -AACTION never gets those :-/A
10:32 < termie> nothing really stopping it, though. It could be done, 'twould just sound ugly and take some bandwidth
10:32 < bunny_home> -hehe
10:32 < termie> but we could pipe a live mp3 stream
10:33 < termie> or vorbis or whatnot
10:33 < termie> i don't know that it could keep up, though
10:33 < termie> if a bunch of people started to talk it could end up days behind
10:33 < bunny_home> -maybe users could select who they want to hear
10:33 < termie> bunny_home: trickier, would require multiple streams... or a global user controlled thing
10:34 < Suw> -ta ra everyone!
10:34 < Suw> -see you later!!
10:34 < bunny_home> -bye suw!
10:34 < yonderboy> +l8r suw
10:34 < termie> by suw :_
10:34 < termie> :)
10:34 < termie> bye
10:34 < termie> ...
10:34 < Suw> -bye bye!! :D
10:34 -!- Suw [~Suw@host81-153-49-138.range81-153.btcentralplus.com] has quit [" Like VS.net's GUI?  Then try HydraIRC ->
          http://www.hydrairc.com <-"]
10:34 < adamhill> -hehe, the voice API for Longhorn is v = new Voice(); v.say(stream or string) :)
10:34  * termie imagines jibot saying "termie beep by suw colon underscore termie beep colon end parens..."
10:35 < bunny_home> -adamhill: is that good or bad?
10:35 < bunny_home> -termie: tts can be a little better than that.
10:36 < adamhill> -we just need to decide on canonical words for smiley faces :)
10:36 < bunny_home> -well
10:36 < bunny_home> -I had to create voice shortcuts for them when I was chatting with dragon
10:36 < bunny_home> -"smiley face"
10:36 < bunny_home> -"winky face"
10:36 -!- talios [~mark@se7en.org] has joined #joiito
10:36 < jibot> -talios is from Auckland, New Zealand and has a blog at http://www.talios.com/
10:36 < bunny_home> -"not sure face"
10:36 < bunny_home> -"sad face"
10:36 < bunny_home> -pretty easy.
10:36 < adamhill> -I was just commenting on how simple Voice API's have come
10:36 < termie> http://joi.ito.com/joiwiki/JiBotVoice
10:36 < bunny_home> -ok adamhill
10:36 < adamhill> -what is o_O?
10:37 < talios> +howdy ho all
10:37 < adamhill> -or ^_^
10:37 < bunny_home> -adamhill: "gaping"?
10:37 < yonderboy> +"smile", no need for the "face"
10:37 < bunny_home> -termie: page looks empty :-)
10:37 < termie> has been updated ;)
10:37 < bunny_home> -yonderboy: how do you know if the person wrote smile or did a :-) then?
10:37 < adamhill> -they are all unique, but naming is always a bitch (as many years of programming have taught me)

10:38 < yonderboy> +bunny_home, tone
10:38 < bunny_home> -hmm ok yonderboy
10:39 < yonderboy> +tts voices have enough range
10:40 < bunny_home> -AACTION is listening to Mere Khwabon Mein Tu by Alka Yagnik & Kumar Sanu from Their Favorites.A
10:40 < bunny_home> -:-)
10:40 -!- bunny_home is now known as bunny_kitchen
10:41 < adamhill> -how would it pronounce 'Suw" or Welsh for that matter
10:41 < bunny_kitchen> -well
10:41 < bunny_kitchen> -jibot needs a multilingual engine then
10:41 < bunny_kitchen> -should do french and hindi too :-)
10:42 < adamhill> -it is very hard to discern roman languages from just characters
10:42 -!- skadz [~skadz@dhcp64-134-213-46.nyrh.nyc.wayport.net] has quit ["Leaving"]
10:42 < adamhill> -non-roman would be lots easier
10:42 < bunny_kitchen> -on peut utiliser des mots-cl&#65533;&#65533;
10:43 < termie> y'all okay with a log of this on the wiki?
10:43 < bunny_kitchen> -speech recognition can recognize languages
10:43 < termie> http://joi.ito.com/joiwiki/JiBotVoice
10:43 < adamhill> -imagine finnish :)
10:43 < bunny_kitchen> -termie: go for it
10:43 < adamhill> -i am
10:43 < bunny_kitchen> -adamhill: the engine just needs to know a few "key" words that allow to identify a language
10:44 < bunny_kitchen> -and which are common enough that chances are they'll appear in a sentence
10:44 < yonderboy> +hmmm, sounds evil
10:44 < adamhill> -'mots dash cee ell tilde A copyright' :)
10:44 < bunny_kitchen> -example: to, in are not in french
10:44 < bunny_kitchen> -the
10:44 < adamhill> -the fallacy is 'key'
10:44 < bunny_kitchen> -what do you mean?
10:44 < adamhill> -there are LOTS of languages
10:44 < bunny_kitchen> -sure
10:44 < bunny_kitchen> -but we know what languages we might see here
10:44 < bunny_kitchen> -and the engine can learn
10:45 < bunny_kitchen> -ie, consider first a language which has already been spoken here
10:45 < adamhill> -maciej is much better at explaining language extraction than me :)
10:45 < adamhill> -oh, yes
10:45 < bunny_kitchen> -:-)
10:45 -!- number5 [~number5@219.137.40.135] has quit [" HydraIRC -> http://www.hydrairc.com <- Nine out of ten l33t h4x0rz prefer it"]
10:45 < bunny_kitchen> -I really need to do my washing-up
10:45 < adamhill> -it is easier to just say X might speak langauge A B or C
10:45 -!- bunny_kitchen is now known as bunny_washing-up
10:45 < bunny_washing-up> -yes, adamhill
10:45 < yonderboy> +browsers fail at it all the time

Interesting info here on making a weblog accessible through VoiceXML (in French): http://www.cybercodeur.net/weblog/commentaires/detailsCarnet.php?idmessage=982 -- bunny