Best way to train the AI for games...
So I've trained a character that I'm semi-happy with, although it took a lot of editing of outside interviews and cutting out the radio transmission voice lines (I assumed that would negatively influence the training). While training a new character, I wondered if splitting audio files would affect the AI training differently/better. Previously, I used a large number of voice lines stitched together in lengthy files. The individual voice lines typically ranged from 3 to 5 seconds of talking and/or yelling and moves on to the next line, all in one file. My question is, would it be better to fully trim each voice line into separate files? If not, is it okay to have normal lines mixed with yelling ones in one file, or could that also affect it negatively? Thanks for your time and insight!
Also to make a voice public can I switch it and save, or do I need to build it again after switching?
I included links to a couple of the sources to give you a better idea of what I'm talking about.
-
It makes no difference how the file is sliced and diced. Get rid of all non verbal lines and anything that is way too distorted (like high pitched yelling different than the regular voice).
Currently you can switch a voice from public to private and back no problem.0
Please sign in to leave a comment.
Comments
1 comment