Hi Sergey,

Sorry.  That section of the web page pointed to the old login  instructions for the FTP site.  It has now been fixed.

thanks,

Ken 

--- (Edited on 4/22/2008 5:26 pm [GMT-0400] by kmaclean) ---

Hi Sergey,

If your file is less than 50 meg, you can also upload it to this forum (I just updated this forum to accept attachments).  Anything greater than 50 meg will not upload ... unfortunately, the CMS I use does not tell you that it has rejected a submission if it is too large.

Ken 

--- (Edited on 4/22/2008 5:32 pm [GMT-0400] by kmaclean) ---

Hi Ken,  Thanks for fixing it. I have uploaded file Sergey-20080422.tgz

Please let me know if there is anything wrong with it.

Thanks,

--Sergey 

--- (Edited on 4/23/2008 9:00 am [GMT-0500] by Sergey) ---

Hi Sergey,

Thanks!

I am travelling right now, so I won't be able to look at it until this Sunday.

Ken 

--- (Edited on 4/23/2008 5:45 pm [GMT-0400] by kmaclean) ---

Hi Sergey,

Was the audio for your submission originally in uncompressed WAV format (or a lossless compressed format like FLAC, etc.), or did you convert it from MP3 (or any other lossy compressed format like OGG, etc.) to wav?

Ken 

--- (Edited on 4/27/2008 10:24 pm [GMT-0400] by kmaclean) ---

Hi Ken,

I Have downloaded from LibriVox in mp3 format.

--Sergey

--- (Edited on 4/28/2008 7:00 am [GMT-0500] by Sergey) ---

Hi Sergey,

Although we prefer the source audio be uncompressed (i.e. WAV) or lossless compressed (i.e. FLAC) audio formats (these give a bit better recognition rates - see this page for some preliminary analysis), I will add your submission to the VoxForge repository, but label it as coming from an MP3 source.

We've got lots of uncompressed LibriVox audio chapters that need processing located here:

Librivox Audiobook Submissions 

The directory needs to be cleaned up, but if you are interested in segmenting more audio, please let me know.  Check the AudioBook Submission Notification Forum to see if a submission has already been processed.

Once we have the Sequitur G2P models trained with the VoxForge Pronunciation dictionary, we should be able to automate most of the segmentation process.

Ken 

--- (Edited on 4/28/2008 3:02 pm [GMT-0400] by kmaclean) ---

Wow! I didn't know you have that much uncompressed audio. That is very impressive.

I am definitely interested in segmenting more audio, time permitting.

Is there any time estimate when Sequitur G2P models will be trained with the VoxForge Pronunciation dictionary?

--Sergey

--- (Edited on 4/29/2008 8:18 am [GMT-0500] by Sergey) ---

Hi Sergeym

>Is there any time estimate when Sequitur G2P models will be trained with the

>VoxForge Pronunciation dictionary?

Another day or two - it has been running since last night on an AMD Athlon  64, 3500 (single core), with 3 Gig, using Fedora FC6.

Ken 

--- (Edited on 4/29/2008 10:18 am [GMT-0400] by kmaclean) ---