Voice Tracking

[ArtisanRadio]

Tune into most internet streams and you’ll hear all music, nary a human (or similar) voice to be heard.  I use internet streams as an example, but Part 15 stations are mostly similar.

Usually, both of these types of stations are run by a single individual, using a computer for automation.  The more adventurous station owners might have the occasional Station ID, jingle or time announcement, but that’s about it.  If you as a listener don’t know the music selections that are being played, you’re out of luck (unless it’s an internet stream and meta data is supplied – many stations don’t even do that).

Voice tracking (the identification of a song or programming selection via voice) is generally left to the ‘big boys’ (i.e., licensed stations), who often don’t use human DJ’s either in this day and age.

Those who are familiar with Artisan Radio also probably know that we developed an automated voice tracking system using computer generated voice (TTS, Text to Speech).  I’ve described the methodology in detail elsewhere.

There are plenty of free TTS engines available, and some great pay ware ones that are relatively inexpensive.

The biggest problem with TTS is robotic sounding voices.  The more natural voices that are available for purchase generally do not allow you to use their output without also purchasing an expensive redistribution license.

Microsoft introduced some reasonably good voices in Windows 8 (Zira, Hazel) and Windows 10 (Eva/Cortana).  The Windows 8 voice audio files can be redistributed (via your Windows license).  Eva can’t be used for TTS unless you go into the Windows registry and move some keys around, so I’m not sure voice audio files using it can be redistributed (the Windows 8 voices can be moved to Windows 7 in a like manner, with the same proviso).  And Windows has now moved their speech services to the cloud (Azure), which requires a subscription for use.  So we’re stuck with what we now have.

Artisan Radio is temporarily offline, but when we come back (to license free FM,
BETS, in Canada) we’re moving to an all talk format, which won’t require voice tracking.  However, I also want to transmit music to myself using AM (the very definition of our license free RSS210, similar to Part 15.219 in the U.S.).

So here was the objective, after some thought.  Burn music tracks with added voice tracking to a DVD (with the player in shuffle mode), which will then feed an AM transmitter (not quite accurate, in between I also have an external compressor, a Symetrix 421M).  Since everything I wanted to listen to is mono (as is the compressor) from sometimes mediocre sources, I was going to compress the tracks down to 64 kbps mono.  This is a perfectly fine bitrate for AM and even FM (try it).  It would also maximize the number of tracks on the DVD (generally between 3000-4000, about a week’s worth transmitting 24/7).

I ended up developing a series of Powershell scripts to, for each track 1) convert to the appropriate bit rate 2) generate the voice tracking audio file in the same bit rate 3) append the two mp3 files and 4) check the integrity of the resulting single track, fixing up headers.

I’d be more than happy to share the scripts if anyone is interested.  Just send me an e-mail through the Forum.

And although I posted the methodology and batch files for automated voice tracking on another site (the former ALPB), I’m sure I can recreate it here if necessary.

[Author]

Posts