2014-05-03, 15:53
ruuk
Finally got around to playing with your speech server. Worked fine once I installed the numpy package.
There's little I can really say other than "it works!". From a users perspective, navigating xbmc with it, it was imperceptible from my perl one. Fast and responsive. If I can figure out how to run yours as a service, I'll ditch the perl one and run yours exclusively.
Oh, actually, there are two very minor things I noticed. It appears the speech server offers ttsd as one of the supported engines. It's a engine that can stream wavs, so I can see how it might end up in that list.
The other minor issue was when I was first configuring the addon to use the python server. After I set perl server to no and selected the engine, it refreshed and perl server was set to yes again and I had to unset that and then choose the voice, and I think I had to unset it again before I went OK, too.
I took a few more minutes and just tested it on FreeBSD. It seemed to work fine. It detected Flite and eSpeak. I only tested flite. It also offered "Google" as an option. I selected it to see what would happen and xbmc immediately crashed as soon as I went OK, which wasn't a surprise. I imagine it offered this option because I have mplayer installed on the host server. If you wanted to offer Google via the speech server, you could probably use mplayer to convert to wav rather than play the mp3 and then deal with it like everything else, though it's probably be better to use mpg123 or sox if available as they're lighterweight if you wanted to go down that path.
But it looks good and those are fairly minor bugs and far from showstoppers. I look forward to seeing how future versions go.
Finally got around to playing with your speech server. Worked fine once I installed the numpy package.
There's little I can really say other than "it works!". From a users perspective, navigating xbmc with it, it was imperceptible from my perl one. Fast and responsive. If I can figure out how to run yours as a service, I'll ditch the perl one and run yours exclusively.
Oh, actually, there are two very minor things I noticed. It appears the speech server offers ttsd as one of the supported engines. It's a engine that can stream wavs, so I can see how it might end up in that list.
The other minor issue was when I was first configuring the addon to use the python server. After I set perl server to no and selected the engine, it refreshed and perl server was set to yes again and I had to unset that and then choose the voice, and I think I had to unset it again before I went OK, too.
I took a few more minutes and just tested it on FreeBSD. It seemed to work fine. It detected Flite and eSpeak. I only tested flite. It also offered "Google" as an option. I selected it to see what would happen and xbmc immediately crashed as soon as I went OK, which wasn't a surprise. I imagine it offered this option because I have mplayer installed on the host server. If you wanted to offer Google via the speech server, you could probably use mplayer to convert to wav rather than play the mp3 and then deal with it like everything else, though it's probably be better to use mpg123 or sox if available as they're lighterweight if you wanted to go down that path.
But it looks good and those are fairly minor bugs and far from showstoppers. I look forward to seeing how future versions go.
