Rockbox.org home
release
dev builds
extras
themes manual
wiki
device status forums
mailing lists
IRC bugs
patches
dev guide



Rockbox mail archive

Subject: Re: speex

Re: speex

From: Jens Arnold <arnold-j_at_t-online.de>
Date: Mon, 19 Nov 2007 23:57:41 +0100

On 19.11.2007, Onj wrote:

> I will have to play with that at some point. I have noticed
> that even though I'm using a 16 K voice for the menus and a 16
> K voice for my voice clips, the voice clips alias a lot more.
> Is the voicebox script slightly different in it's
> construction, from the stuff under say, Cygwin? I'm not sure
> why it sounds the way it does, but I'm trying to track down
> the problem.
> Even when I used the same voice to make talk clips as I did
> for the menus with -q 10 and -c 10, the talk clips aliased but
> the menu voice did not.

What SAPI engine are you using? SAPI engines can produce their
output at different sample rates. If that setting isn't optimal
for the engine it will produce aliasing. We cannot change this
behaviour fundamentally. The fix is to select the optimal format
for the engine used.

The sapi_voice.vbs used by voice.pl is clever enough to detect
the engine, and set the optimal SAPI output format (for engines
we know). voiceBox, however, does not, but always uses the
format set near the top of voiceBox.wsf (const AUDIO_FORMAT).
This is something I did not change - it has always been like
this. I did not want to rewrite the whole voiceBox thing but
just wanted to make it usable for speex clips.

You can change the format there to match the engine you use, or
step ahead and make voiceBox more intelligent. But maybe rbutil
is the way to go (once that learned to produce speex clips).

Regards, Jens
Received on 2007-11-19


Page was last modified "Jan 10 2012" The Rockbox Crew
aaa