User avatar
Stewart Russell @scruss@xoxo.zone
1w
DECtalk babbling ...

These sort-of sound like English words. I trained the smallest possible language model (a Markov chain of length 2) on the phonemes generated when DECtalk read a chapter of a book. The model picks the next sound as a random one likely to follow the previous ones.
5
1
0
0
1w
@scruss @jaybird110127 A more natural sounding voice and you’d have a decent asmr generator.
1
0
0
0

User avatar
Stewart Russell @scruss@xoxo.zone
1w
@fastfinge @jaybird110127 I like my voices scrungly, though
0
0
0
0