[Freeswitch-users] Proper prompt gain/level

[Bryan Smart]

Hi, Michael.

Thanks for posting this.

Have you considered using the higher quality sample rate conversion features in new versions of sox?

A starting place for info is here:
http://sox.sourceforge.net/SoX/Resampling

Perhaps change:

sox -v 0.2 $base_dir/$dir/$filename -r $rate -c 1 $tmp_dir/$voice_dir/$dir/$rate/$filename

to:

sox $base_dir/$dir/$filename -c 1 $tmp_dir/$voice_dir/$dir/$rate/$filename rate -v -I $rate dither

rate -v -I $rate: Replaces -r $rate. Uses the newer "very high quality" mode (instead of "high quality") with Intermediate Phase response (instead of linear). This setting is superior to the default, but slower to process.

dither: Applies basic triangular dithering. There are other dithering strategies, but, based on what I've read and experienced, they aren't as useful when down-sampling to low rates.

Might produce slightly higher quality files, the next time it is necessary to regenerate them.

Bryan

On Jun 28, 2011, at 11:18 AM, Michael Collins wrote:

You can download the 48kHz files from files.freeswitch.org<http://files.freeswitch.org/>. The script itself is nothing special - all it does is cycle through the target sampling rates and run each file through sox. Here's a copy of the script:

base_dir="48000"
rates="48000 32000 16000 8000"
version="1.0.16"
voice="en-us-callie"
voice_dir="en/us/callie"
tar_path="../../.."
tmp_dir="tmp"
CWD=`pwd`

for rate in $rates; do \
  mkdir -p $tmp_dir/$voice_dir
  for dir in `ls $base_dir`; do \
    test -d $tmp_dir/$voice_dir/$dir/$rate || mkdir -p $tmp_dir/$voice_dir/$dir/$rate; \
    for filename in `ls $base_dir/$dir`; do \
      echo sox -v 0.2 $base_dir/$dir/$filename -r $rate -c 1 $tmp_dir/$voice_dir/$dir/$rate/$filename; \
      sox -v 0.2 $base_dir/$dir/$filename -r $rate -c 1 $tmp_dir/$voice_dir/$dir/$rate/$filename; \
    done ; \
  done ; \
  cd $tmp_dir
  echo tar -cvzf $CWD/$tar_path/freeswitch-sounds-$voice-$rate-$version.tar.gz *; \
  tar -cvzf $CWD/$tar_path/freeswitch-sounds-$voice-$rate-$version.tar.gz *; \
  cd -
  rm -rf $tmp_dir
done

cd $tar_path
for rate in $rates; do \
  openssl dgst -sha1 freeswitch-sounds-$voice-$rate-$version.tar.gz > freeswitch-sounds-$voice-$rate-$version.tar.gz.sha1 ; \
  openssl dgst -md5 freeswitch-sounds-$voice-$rate-$version.tar.gz > freeswitch-sounds-$voice-$rate-$version.tar.gz.md5 ; \
done
cd $CWD

Let me know if you have any suggestions.
-MC

On Mon, Jun 27, 2011 at 8:25 PM, Bryan Smart <bryansmart at bryansmart.com<mailto:bryansmart at bryansmart.com>> wrote:
-v is a multiplier. So, -v 0.2 reduces gain to 20% of original.

It is typical to supply professionally recorded audio with peaks boosted to 0DBFS. That is probably how the prompts were originally supplied. It makes since that -v 0.2 would reduce the max power to 20%, around -15 to -16 DBFS.

The FS build process downloads pre-processed versions of these files. Is there somewhere online that I can find the source recordings+script that produces the pre-processed files that are downloaded by the build process?

Bryan

On Jun 27, 2011, at 7:32 PM, Michael Collins wrote:



On Mon, Jun 27, 2011 at 3:07 PM, Anthony Minessale <anthony.minessale at gmail.com<mailto:anthony.minessale at gmail.com>> wrote:
Have a look at the 48khz versions of the files, those should be the originals.
Someone runs a batch sox command on them to get them to the other levels.

Just for the record, the sox command used to normalize the sounds is:
sox -v 0.2 $file_in -r $rate -c 1 $file_out

The -v format is the "volume" format and if I read the sox man page correctly it means a "linear amplitude adjustment". What I can't tell from the man page is whether .2 means "reduce by 20%" or "reduce to 20%" or something else. In any case, like Tony says, we are not picky about this as long as it's not pointlessly loud.

-MC
_______________________________________________
Join us at ClueCon 2011, Aug 9-11, Chicago
http://www.cluecon.com<http://www.cluecon.com/> 877-7-4ACLUE

FreeSWITCH-users mailing list
FreeSWITCH-users at lists.freeswitch.org<mailto:FreeSWITCH-users at lists.freeswitch.org>
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org<http://www.freeswitch.org/>


_______________________________________________
Join us at ClueCon 2011, Aug 9-11, Chicago
http://www.cluecon.com<http://www.cluecon.com/> 877-7-4ACLUE

FreeSWITCH-users mailing list
FreeSWITCH-users at lists.freeswitch.org<mailto:FreeSWITCH-users at lists.freeswitch.org>
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org<http://www.freeswitch.org/>


_______________________________________________
Join us at ClueCon 2011, Aug 9-11, Chicago
http://www.cluecon.com 877-7-4ACLUE

FreeSWITCH-users mailing list
FreeSWITCH-users at lists.freeswitch.org<mailto:FreeSWITCH-users at lists.freeswitch.org>
http://lists.freeswitch.org/mailman/listinfo/freeswitch-users
UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users
http://www.freeswitch.org

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.freeswitch.org/pipermail/freeswitch-users/attachments/20110628/c1adb9ef/attachment-0001.html