# machine-learning
- a
Ask For Dataset = Navy Seal
08/18/2021, 8:48 PM@User is 170 epoch ok? - h
hecko
08/18/2021, 8:48 PMprobably yeah - a
Ask For Dataset = Navy Seal
08/18/2021, 8:53 PM@User should i change the batch size? - a
Ask For Dataset = Navy Seal
08/18/2021, 8:53 PMit's 14 - m
mega b
08/18/2021, 8:54 PM> @User I trained the tacotron for 15 minutes, and got this a minute or so before it ended. Is this spectrogram good? @Ask For Dataset = Navy Seal keep training - m
mega b
08/18/2021, 8:54 PMdont change the batch size - a
Ask For Dataset = Navy Seal
08/18/2021, 8:54 PMok - m
mega b
08/18/2021, 8:54 PMCould you send a screenshot of the graph here? - m
mega b
08/18/2021, 8:55 PMWait hmm - a
Ask For Dataset = Navy Seal
08/18/2021, 8:55 PMit just started training - m
mega b
08/18/2021, 8:55 PMcould you send your transcript here - a
Ask For Dataset = Navy Seal
08/18/2021, 10:08 PM@User same result - h
hecko
08/18/2021, 10:08 PMm - h
hecko
08/18/2021, 10:08 PMgenuinely no clue - h
hecko
08/18/2021, 10:09 PMtriple-check the format maybe??? - h
hecko
08/18/2021, 10:09 PMlike maybe you're doing 24-bit wav - h
hecko
08/18/2021, 10:10 PMor somehow 16-bit big-endian (is that even a thing) - a
Ask For Dataset = Navy Seal
08/18/2021, 10:12 PMno i have signed 16 bit PCM - m
mega b
08/18/2021, 10:27 PMkeep an eye out on audio longer than 12 seconds - m
mega b
08/18/2021, 10:27 PMand shorter than 1.5 - h
hecko
08/19/2021, 7:57 AMlistening to the samples on this page and thinking "wow this gt ai is so realistic why i haven't i heard of it" https://speechresearch.github.io/fastspeech2/ - h
hecko
08/19/2021, 7:57 AM(gt = ground truth = human) - m
mega b
08/19/2021, 2:39 PMthis would be awesome to try - m
- m
matty1406
08/19/2021, 5:32 PMMy Dr Trager model... - u
user
08/20/2021, 12:52 AM - u
user
08/20/2021, 12:52 AMthere's a pytorch version - h
hecko
08/20/2021, 12:53 AMthis has it too https://github.com/TensorSpeech/TensorFlowTTS - u
user
08/20/2021, 12:56 AM