If I use Microsoft's Text-to-Speech engine on text whose copyright I own, who owns the rights to the resulting audio file? Do I own it in every instance, or are there restrictions of the IP rights I have over the resulting audio files?
UPDATE I had not noticed that the original is from 2010. I will also try to include a more modern citation: The Microsoft license (from what I can find) appears to have not changed from what I can see.
Current Microsoft Services agreement appears to continue what was said before. The best way to test that is to download it and see what pops up on the screen.
Cepstral Personal voices are for personal use only and are NOT licensed for audio distribution. This means the audio you create is for your use only and cannot be shared with others or used in videos, presentations, or webpages. If you are interested in an audio distribution license, please contact sales.
This summary from 2013 points to some programs.
This thread form 2014 gives prices for a number of synthesizers with quotes from email responses from the distributors.
Text-to-Speech Costs – Licensing and Pricing points out that the use of the voices itself would require a license and pricing. As a result, you would need to see how they explicitly limit the use of the voices within the program.
I did find this reference Commercial use of voices
If you use the free Microsoft voices there are no restrictions, but with any of the premium voices like AT&T Natural Voices, neospeech and Cepstral voices, commercial licensing would be required for any type of commercial use or ability to distribute the audio files.
With the AT&T Voices there are severe restrictions, with a couple of license types we can offer starting at $5000.
With Neospeech voices we have a little more flexibility and pricing starts at $1500.
With Cepstral voices prices start at $199.
With the Cepstral thing, you have to buy the voices you want to use ($29.95 each), then it is $199/year for right to distribute the audio files from those voices. Not $199 per voice, just $199. You can buy that at http://www.nextup.com/redir/cepstral.html
The license statement for Cepstral is
"You are about to purchase Cepstral's Audio Distribution License (ADL)
This license will provide rights to distribute audio files generated using the Cepstral engine and voices. For example, this license is appropriate for persons or companies who wish to make audio available via HTML, Radio, FTP, CDROM, Flash Animation, Virtual Agents, Recorded Phone Messages, or other distribution means. The term audio covers any file compression algorithm or container format such as WAV, MP3, etc.
The ADL is required to distribute audio regardless of whether your content is offered for a fee or not.
We hope that by making the ADL available to our customers over the Internet, that we can encourage creative uses and applications for our high-quality text to speech voices.
Thank you for your patronage. If you have further questions, please contact email@example.com. "
FYI: On MacOS X, the voices that come with the operating system or that can be downloaded for free have a license that allows you private use, no distribution would be allowed. Since the voices are usually hundreds of megabytes of actually recorded speech, it seems to make sense that there is copyright in derived works (for example, a text that you converted to sound using these sounds).
The copyright of the converted text would be shared between you and the copyright holder of the voices. So you would need a license for distribution. In this case, storing the recording on your computer or your own phone for playback would be legal because you have a license, but distribution wouldn't.