# machine-learning
- u
user
05/19/2021, 4:06 AMThat's only half the dataset, I still need to divide and transcribe the rest of the interview so I can't wait for that - h
HoodyP
05/19/2021, 7:23 AMAfter nearly 24 hours, 706 epochs, 15545 iterations and using 3 Google accounts, my Clank voice model has finally reached my validation loss goal of around the range of 0.06-0.05! Time to test it out and will upload the model + dataset soon - p
pi
05/19/2021, 7:35 AMhow large is your dataset? - h
HoodyP
05/19/2021, 7:37 AMAbout 680 files of lines - h
HoodyP
05/19/2021, 7:38 AM42 mins total - h
HoodyP
05/19/2021, 7:38 AMI just tested and most of the sample is just fine! - h
HoodyP
05/19/2021, 7:38 AMIt’s really good! - h
HoodyP
05/19/2021, 7:39 AMI’m a bit bummed out cuz even with alot of lines, my Clank model doesnt support Arpabet 😦 - h
HoodyP
05/19/2021, 7:40 AMBut still, its actually quite better than the Ratchet one I did! - p
pi
05/19/2021, 7:40 AMthat's not how that works - h
HoodyP
05/19/2021, 7:40 AMHuh? - p
pi
05/19/2021, 7:40 AMyou need to convert your transcripts into arpabet before training - p
pi
05/19/2021, 7:40 AMif you want arpabet support - h
HoodyP
05/19/2021, 7:41 AMOh... - h
HoodyP
05/19/2021, 7:41 AMThen maybe I should train it again soon with an Arpabet transcript - h
HoodyP
05/19/2021, 7:42 AMAnyways, it’s WAAAY better than before! I’m gonna wait till my Colab cool off before I retrain it again with Arpabet! - h
HoodyP
05/19/2021, 7:43 AMHowever I will post the model I have made, before I update it soon, or some help after I posted the dataset - u
user
05/19/2021, 4:43 PMIs there someone who would like to collaborate on a training? i have access to studio quality rap acapellas with no effects on them. Feel free to contact me to speak 🙂 - o
OctolingVladTTS
05/19/2021, 5:06 PMI found a way to remove the background music without splitter.ai - o
OctolingVladTTS
05/19/2021, 5:06 PM - s
SidPlays_144p
05/19/2021, 5:09 PMi'm not sure what that has to do with his comment - a
Awesome Face (GTX 1660 Ti)
05/19/2021, 7:13 PMarpabet is this: {HH AH L OW . W ER L D . ? .} - a
Awesome Face (GTX 1660 Ti)
05/19/2021, 7:13 PMnot like, vocal removal smh - a
Awesome Face (GTX 1660 Ti)
05/19/2021, 7:13 PMit's useful to fix pronounces - k
kitsuner
05/19/2021, 7:29 PMYou don’t need it too much though. I’ve managed to imitate sounds pretty easily in other ways - k
kitsuner
05/19/2021, 7:30 PMThe problem is the AI doesn’t understand breaks or if words have U’s in them it says them too fast - c
ccateni
05/19/2021, 7:30 PMWhich of the sites are better? There are so many background removal sites now... - c
ccateni
05/19/2021, 7:30 PMThats the bigger problem. - k
kitsuner
05/19/2021, 7:32 PMLike trebuchet could be treh bue shey - k
kitsuner
05/19/2021, 7:32 PMAnd the AI might get it right