Fwd: Integrating speech speed control

[Bill Cox]

I'm resending this e-mail since the attachments were too large and
were rejected by the mail server.  I've removed the original natural
speech sample, but all the others are still here.

---------- Forwarded message ----------
From: Bill Cox <address@hidden>
To: address@hidden
Date: Wed, Nov 3, 2010 at 2:51 PM
Subject: Integrating speech speed control
To: speechd at lists.freebsoft.org


I've written a package I'm calling sonic, which is a very simple ANSI
C GPL program for speeding up or slow down speech. ?Distortion is very
low, and the results even at 6X speed up are good. ?I've attached some
samples that show how it performs.

I would be happy to help integrate it into speech dispatcher if it
makes sense. ?As an example of it's use, espeak tops out at about 2.6X
speed up, when using -s 450. ?libsonic allows it to go far faster than
that, and even at speed 2.7X speed up, I believe libsonic creates
better sounding speech by speeding up default speed espeak than espeak
does natively. ?Some people use voxin purely for the enhanced
productivity possible due to it's ability to generate higher speed
speech than any available open-source alternative. ?With libsonic
integration into speech-dispatcher, I think just about any speech
synthesiser could run at any speed desired.

The first two files, talking.ogg and talking_2x.ogg are samples of my
dad talking. ?The first is the original file and the second is sped up
with sonic by 2X. ?The other files show how sonic performs at speeding
up synthesized speech. ?espeak_s450.ogg was generated by espeak, using
-s 450. ?espeak_sonic.ogg was generated by having espeak synthesize at
it's default rate, and then speeding it up using sonic by a factor of
2.6. ?These files are about the same speed, but sound quite different.
?voxin_s110.ogg is voxin using the say program and -s 110.
voxin_sonic.ogg is voxin at default speed, then sped up by sonic by
3.1X.

Source code for sonic can be downloaded from the Vinux Lucid PPA at
launchpad.net. ?Is there any interest?

Regards,
Bill
-------------- next part --------------
A non-text attachment was scrubbed...
Name: talking_2x.ogg
Type: audio/ogg
Size: 372472 bytes
Desc: not available
URL: 
<http://lists.freebsoft.org/pipermail/speechd/attachments/20101103/9f092a62/attachment-0005.ogg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: espeak_s450.ogg
Type: audio/ogg
Size: 89778 bytes
Desc: not available
URL: 
<http://lists.freebsoft.org/pipermail/speechd/attachments/20101103/9f092a62/attachment-0006.ogg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: espeak_sonic.ogg
Type: audio/ogg
Size: 81130 bytes
Desc: not available
URL: 
<http://lists.freebsoft.org/pipermail/speechd/attachments/20101103/9f092a62/attachment-0007.ogg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: voxin_s110.ogg
Type: audio/ogg
Size: 31282 bytes
Desc: not available
URL: 
<http://lists.freebsoft.org/pipermail/speechd/attachments/20101103/9f092a62/attachment-0008.ogg>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: voxin_sonic.ogg
Type: audio/ogg
Size: 31432 bytes
Desc: not available
URL: 
<http://lists.freebsoft.org/pipermail/speechd/attachments/20101103/9f092a62/attachment-0009.ogg>