rtitmuss;259270 Wrote: 
> 
> This sounds like an interest project and way to use Jive. I think it
> would be pretty easy to extend the Jive software to interface with a
> text-to-speech engine. The hard part would be getting the
> text-to-speech engine running on the embedded processor. Do you know
> what the cpu and ram requirements for a linux text-to-speech system
> are?
> 

Why not run the TTS engine on the SC7 side? Use Festival or whatever to
create the audio files, and have Jive fetch & cache those audio files.

A lot of text in SqueezeCenter is localized in "strings.txt" files, but
it is also fairly common for applications to combine one or more
localized strings with information from elsewhere -- the database of
music information, for instance. In some situations like browsing the
local music collection, it's entirely non-localized data. So it would
be nice if Jive could pre-fetch TTS files much like the suggestion made
by sighted users that Jive should pre-fetch and cache album art images.
It should be helpful that the Controller includes an SD card slot, as
that would give you the ability to store a lot of TTS snippets.

Jeronimo, have you used Rockbox? It has spoken interface features, and
I've recommended it as a model for other UI issues, but I don't know
how "good" its spoken UI really is. It sounds like the basic Rockbox
spoken UI model is much like the old SlimServer "cover.jpg" album art
model -- Rockbox users make separate MP3 files for each spoken string
and put them in their player/music directory. 
http://www.rockbox.org/twiki/bin/view/Main/BlindUsersIndex
http://www.rockbox.org/twiki/bin/view/Main/BlindFAQ#How_do_I_let_Rockbox_speak_direc
I don't think that would be as good as on-demand TTS, especially now
that localized SqueezeCenter strings can be printf-style format
strings, but it would probably be simpler to code. And cache -- if
you're pre-generating the spoken UI sound files, you could copy them to
the Controller's SD card yourself, and all Jive would have to know is
how to determine the path/filename for a particular string, check the
local FS, and use the file if it's there. That could be as simple as
hashed MD5 names, e.g. the MD5 of "Music Library" is a 32 character hex
string starting with "08b6d", so Jive might look for
/mnt/mmc/spoken-ui/0/8/b/08b6d etc. .mp3 as a spoken version of that
string.

-Peter


-- 
peterw

http://www.tux.org/~peterw/
free plugins: http://www.tux.org/~peterw/#slim
AllQuiet BlankSaver BottleRocket FuzzyTime SaverSwitcher SleepFade
StatusFirst SyncOptions VolumeLock
------------------------------------------------------------------------
peterw's Profile: http://forums.slimdevices.com/member.php?userid=2107
View this thread: http://forums.slimdevices.com/showthread.php?t=42297

_______________________________________________
jive mailing list
[email protected]
http://lists.slimdevices.com/cgi-bin/mailman/listinfo/jive

Reply via email to