# machine-learning
- e
Emoto
02/02/2022, 1:35 PMMekatron was updated in November tho, is it really dead? - e
Emoto
02/02/2022, 1:38 PMI send them a tweet to see if they give me a somewhat decent reply that I can translate into English lol - e
Emoto
02/02/2022, 1:42 PMWas there any update on getting Tacotron2 to work/fork with the World Vocoder? Because it would be quite a blast to have fast wav renders when you wanna test something out and don't want to train a seperate vocoder model on talknet or hifigan lol - e
Emoto
02/02/2022, 1:47 PMHm. Actually seems like the world vocoder would struggle without some sort of f0 from the tacotron2 NN - e
Emoto
02/02/2022, 1:47 PM - e
Emoto
02/02/2022, 1:48 PMOh yeah now I see why people weren't able to use jt - e
Emoto
02/02/2022, 1:48 PMCopy codeTo validate our design choices, we present ablation studies of key components of our system and evaluate the impact of using mel spectrograms as the input to WaveNet instead of linguistic, duration, and F0 features. - e
Emoto
02/02/2022, 1:48 PMMay as well shelf that project - t
The Wiggles and High School D×D
02/02/2022, 10:35 PMSo I'm looking through anything about LibriSpeech metadata after downloading the LibriSpeech dataset Because it appears I would need my metadata to be prepared in the "LibriSpeech format" - t
The Wiggles and High School D×D
02/02/2022, 10:36 PMI'm not sure what I'm meant to be looking for All I can find are files labelled SPEAKERS.txt and CHAPTERS.txt that call themselves metadata of the audio collection although aren't themselves called metadata.csv - h
haru0l
02/03/2022, 12:16 PMgo deeper - u
user
02/03/2022, 12:51 PMWhy the Brazilian voices are fading from uberduck?!! - t
The Wiggles and High School D×D
02/03/2022, 3:22 PMI saw that but I wasn't sure if it counted? - t
The Wiggles and High School D×D
02/03/2022, 3:22 PMIf it does I guess I have to figure out how to reformat it all - s
skyglowfading
02/03/2022, 3:47 PMMade my first ever tacotron model to test it out! I used Glamrock Freddy. This is the result. - u
user
02/03/2022, 3:49 PMIt's a shame that Glamrock Freddy isn't allowed on uberduck because Kellen Goff doesn't want his voice to be cloned. - u
(Dawn) Will Draw Fictional Women
02/03/2022, 3:50 PMthe banned voice list should really be put in like #813143605862203442 or something - s
skyglowfading
02/03/2022, 3:51 PMAh, damn. Wasn't planning on uploading it anyway but this will be great for personal use! - s
skyglowfading
02/03/2022, 3:51 PMJust testing out the features anyway - t
The Wiggles and High School D×D
02/03/2022, 5:04 PMHe is generous enough to provide custom voice clips of his own - t
The Wiggles and High School D×D
02/03/2022, 5:04 PMI guess you could resort to cutting and stitching existing dialogue - i
IBob012
02/03/2022, 7:02 PMsome were removed because of the possibility of the voice actors not liking it theres more on falatron.com tho - b
Brigadier
02/03/2022, 11:56 PMHow do you tell if the graph is any good? This is mine - g
Gosmokeless28
02/03/2022, 11:57 PMLooks fine to me. - b
Brigadier
02/03/2022, 11:58 PMWhat do they actually mean tho? - b
Brigadier
02/03/2022, 11:58 PMAnd I did just pick a particular one for instance one of the next ones was this - g
Gosmokeless28
02/03/2022, 11:59 PMLol, that's something I'm not that sure about myself - c
Couch
02/03/2022, 11:59 PMgraph represents attention - c
Couch
02/04/2022, 12:00 AMalso a small graph like that means one of your clips are about a second long - c
Couch
02/04/2022, 12:00 AMdecoder time step is in centi second