My reply:

Hi Mike,

Thanks for thinking of VoxForge!

Unfortunately, I think we might have some problems with Copyright. 

I am not sure what kind of licence the AIRS-LA might have with Motor Trend, but I'm guessing that it would not likeley extend to relicensing them under GPL (or compatible license) and subitting those recordings to VoxForge for any purpose (including commercial use) under the GPL...

If that is not the case, then please let me know,

thanks,

Ken

--- (Edited on 11/25/2009 1:26 pm [GMT-0500] by kmaclean) ---

Ken,

                To be honest, I don’t think that there is any sort of licensing agreement with MT, or any other magazine that’s read for that matter.  My guess is that because it’s the articles aren’t being re-sold at all, but rather being provided as a service to the blind and sight-impared community.  Also, from reading the VoxForge site, I got the impression that the audio samples people provided were used ‘internally’ (for lack of a better way of saying it) so that you could better train the engine to recognize a broad spectrum of voices and inflections.  Is that correct, or did I mis-read the usage of the samples?

                If the samples are used only ‘internally’, I can’t see there being any issue.  However, if they’re going to be disseminated to an outside developer community, I can see how things might get a little sticky.  Let me know how they’re going to be used and I’ll check with my contact over at AIRS-LA to see if he can forsee any issues from the legal standpoint.  Worse comes to worse, I’ll track down someone at Motor Trend and see if they’ve got any issues with it.

                 Let me know when you can and we’ll go from there.

Thanks,

-Mike

--- (Edited on 11/25/2009 1:27 pm [GMT-0500] by kmaclean) ---

Hi Mike,

The whole idea of creating GPL licenced acoustic models was to ensure that the 'source' audio was made available for others to generate their own acoustic models (providing the 'source' code to a user of a program is the essence of the GPL license itself). 

Otherwise, we could do as you suggest and scrape audio from anywhere and everywhere (TV, Podcasts, etc.) to generate acoustic models. 

Copyright protects the form of expression of a work, and any derivative works, including the 'reading' of those works.  So I doubt that commercial publications would be interested in letting their IP go... 

If you are interested, what you might do is contribute a reading to LibriVox, which uses project Gutenberg texts (out-of-copyright books) to create public domain audio books.  You could find a title of interest to your blind listeners, record it to LibriVox (in MP3 or Ogg) and to VoxForge (in WAV or FLAC)... two for one  :)

thanks for looking for ways to help out,

Ken

--- (Edited on 11/25/2009 1:27 pm [GMT-0500] by kmaclean) ---

Ken,

                Okay – now I understand.  Based on that description (and hence the reality of things), I’d say you’re right.

                I’ll have to do some digging into LibriVox over the Holiday and see what they’ve got as far as texts – it sounds like it could be a lot of fun.  J

Thanks for the information and have a Happy Thanksgiving!

-Mike

--- (Edited on 11/25/2009 1:28 pm [GMT-0500] by kmaclean) ---