4 ms·
ASMR is probably pretty well suited to ML. I had a sox one liner generating ~infinite content from [no talking] ASMR videos, which worked surprisingly well with
by a_imho 8y ago
ASMR is probably pretty well suited to ML. I had a sox one liner generating ~infinite content from [no talking] ASMR videos, which worked surprisingly well with some post processing in audacity. I would guess today some services could synthesize convincing speech too.
- wrinkl3 8y agoWould you mind sharing the said oneliner? I've never worked with sox, but it looks interesting.
- a_imho 8y agoSorry, it seems I lost it. It was nothing fancy at all though, I simply piped youtube-dl to sox, cut up the original audio on 'silence' (some tweaking the thresholds for what is considered silence was necessary) then concatenated the resulting segments in random order until it was long enough. Then some manual touches with audacity. End result was something like this https://www.youtube.com/watch?v=eO_Z3MLZrNQ https://www.youtube.com/watch?v=eO_Z3MLZrNQ