Re: Making a speech synthesizer
You can get access to a high quality recording setup and someone who can do a bunch of sound editing, spend weeks or so making the samples that something like Festival requires, figure out the pronunciation rules for whatever language you want, and read the necessary papers to figure out how to run the Festival toolchain. Unfortunately figuring out pronunciation rules is really hard and this won't be super high quality or a fast synth.
Alternatively you can do roughly the same thing with Espeak, but in that case you have to manually edit everything by hand because Espeak's method of synthesis doesn't work off recordings. For that you'll probably need to write some custom scripts of some sort. Sighted people can figure out starting points by looking at spectragrams, but we can't, so you'd have to make up for that.
But in either case the biggest blocker is probably that you need some sort of linguistics person to figure out how to do the pronunciation rules, because outside of one of Google's machine learning things, no one's managed to automate that.
it's technically doable but it's a 6 month project for multiple people to have anything high quality, if I had to estimate. Maybe a little less if someone already has the expertise and you don't have to learn anything. Everyone involved will need to be fluent in the language it's for as well, so for the rarer languages you'll have trouble getting a team together.
-- Audiogames-reflector mailing list Audiogames-reflector@sabahattin-gucukoglu.com https://sabahattin-gucukoglu.com/cgi-bin/mailman/listinfo/audiogames-reflector
