# machine-learning
- y
yousef
01/15/2022, 4:52 AMYo (dawn) - y
yousef
01/15/2022, 4:52 AMi have 346 wavs - y
yousef
01/15/2022, 4:52 AMwhat do i use for the batch size? - m
mega b
01/15/2022, 5:16 AM14 - u
(Dawn) Will Draw Fictional Women
01/15/2022, 8:05 AMwell look at that auto transcriber is public now - m
mega b
01/15/2022, 5:21 PMvery nice 👍 - m
mega b
01/15/2022, 5:22 PMYou should try making your own reference audio and read a script - j
jens
01/15/2022, 5:36 PMIm currently trying to train a new one because the first one had pretty high loss values still and was more of an expirement - j
jens
01/15/2022, 5:39 PMIm quite surprised with the training times compared to other software like dfl (ofcourse image deepfaking costs way more resources) but its still pretty fast tbh - y
yousef
01/15/2022, 8:04 PMhttps://uberduck.ai/#voice=cypher-valorant&mode=tts-basic cypher from valorant that me and @e.e did - e
e.e
01/15/2022, 8:05 PMYeah, It already added you and me into the contribute list. Don't worry. 🙂 - y
yousef
01/15/2022, 8:05 PMOh i know - y
yousef
01/15/2022, 8:05 PMJust showcasing it - y
yousef
01/15/2022, 8:05 PMFor people to use - m
matty1406
01/15/2022, 10:40 PMthe potassium was too much for afton lol - u
(Dawn) Will Draw Fictional Women
01/15/2022, 10:43 PMhigh validation loss moment - m
matty1406
01/15/2022, 10:44 PM0.11 - u
(Dawn) Will Draw Fictional Women
01/15/2022, 10:44 PMi rest my case - u
(Dawn) Will Draw Fictional Women
01/15/2022, 10:44 PM>not sub-0.095 - m
matty1406
01/15/2022, 10:44 PMit was just that one sentence that stroked - m
matty1406
01/15/2022, 10:45 PMit still works really well with long sentences - m
matty1406
01/15/2022, 10:45 PMalso it got to 250 before 0.09 - u
(Dawn) Will Draw Fictional Women
01/15/2022, 10:45 PM>doesnt stop training manually - p
Pekka
01/16/2022, 3:46 PMwhat should the talknet spectrogram generator training validation loss be? - u
(Dawn) Will Draw Fictional Women
01/16/2022, 5:37 PMdepends