Brandon Thomas
af80527c47
Check TT2 model for validity and store file size
2021-06-07 22:56:56 -04:00
Brandon Thomas
dfc52681cd
calculate audio metadata
2021-06-05 15:47:21 -04:00
Brandon Thomas
9c1a1e88f9
simplify mel generation
2021-06-05 15:36:30 -04:00
Brandon Thomas
f804d0216b
some form of spectrogram
2021-06-05 14:13:31 -04:00
Brandon Thomas
000f3b9ef5
put spectrogram into correct dimensions
2021-06-04 03:58:17 -04:00
Brandon Thomas
ec450230f8
saves spectrogram json and uploads it!
2021-06-04 03:54:02 -04:00
Brandon Thomas
716a75c4f4
THAT DID THE TRICK! I was using the wrong wav encoding!
2021-06-02 22:35:08 -04:00
Brandon Thomas
a1c77fb527
Copying waveglow files @ 2fd4e63 (per collab synthesis.ipynb)
2021-06-02 22:26:55 -04:00
Brandon Thomas
7744f0e324
working on adapting the waveglow/inference to sound non-noisy
2021-06-02 22:17:27 -04:00
Brandon Thomas
077a4239d8
sounds better, but still bad
2021-06-02 21:53:30 -04:00
Brandon Thomas
ae91e19b9f
fixup filename, debugging
2021-06-02 21:15:33 -04:00
Brandon Thomas
dadb26c5f2
path bugfixing (still not working)
2021-06-02 00:28:54 -04:00
Brandon Thomas
46f878beaa
tts inference script
2021-06-01 16:21:03 -04:00
Brandon Thomas
05c9eff1c5
TTS inference queries+codegen, job/script WIP
2021-06-01 02:03:52 -04:00
Brandon Thomas
164cb46fff
add debug print to tts inference script
2021-05-31 23:42:29 -04:00
Brandon Thomas
335d355d64
add w2l deps, update readme
2021-05-31 21:45:45 -04:00
Brandon Thomas
e31a5557b6
add working requirements
2021-05-31 18:29:03 -04:00
Brandon Thomas
5f77848304
wav inference works e2e! It's got staticy output, works!
2021-05-31 17:58:21 -04:00
Brandon Thomas
f23ceb22bf
add waveglow @ master
2021-05-31 17:53:42 -04:00
Brandon Thomas
79f7c6c82d
mel histograms look good! (this is just a test)
2021-05-31 17:38:56 -04:00
Brandon Thomas
49dde0a2ff
Add utility from my old experimentation (details)
...
~/dev/3rd/nvidia-tacotron2
commit f6be28d1f89b5e0bae89a771c5d116722f8661cb
Author: Brandon Thomas <echelon@gmail.com >
Date: Sat Jan 11 01:24:26 2020 -0500
Running on CPU, also wav images
2021-05-31 17:36:43 -04:00
Brandon Thomas
04bdfa77fb
spectrogram inference works !
2021-05-31 17:33:09 -04:00
Brandon Thomas
34940f75b1
inference wip
2021-05-31 17:26:32 -04:00
Brandon Thomas
14f1fd327c
requirements from halide that seem to be almost working
2021-05-31 17:21:55 -04:00
Brandon Thomas
21c2194828
additional comment
2021-05-31 16:53:25 -04:00
Brandon Thomas
a423cf6de9
a bunch of alternative requirements files
2021-05-31 16:52:41 -04:00
Brandon Thomas
a05b1ed551
revert Wav2Lip using Python3.6 with a PPA hack
2021-05-30 14:41:39 -04:00
Brandon Thomas
eee60e3cdd
upgrade python package
2021-05-30 13:44:43 -04:00
Brandon Thomas
41fc0ee1ec
fix requirements?
2021-05-30 06:45:57 -04:00
Brandon Thomas
943047e59c
an old "inference" script I found seems to work :whatever:
2021-05-30 06:09:33 -04:00
Brandon Thomas
ecc2c3bec6
I installed my own requirements.txt somehow not realizing it was there
...
what
2021-05-30 04:55:49 -04:00
Brandon Thomas
f5f5a8dd93
add tacotron2
2021-05-30 04:27:51 -04:00
Brandon Thomas
ece7d66e58
end bump
2021-05-29 15:12:11 -04:00
Brandon Thomas
2d7f46cc57
progress, but broke one of the jobs
2021-05-29 13:58:06 -04:00
Brandon Thomas
a8211a1f9b
this fixes duration metadata
2021-05-29 13:20:50 -04:00
Brandon Thomas
9075f97d7a
output metadata (wrongish?)
2021-05-29 13:03:19 -04:00
Brandon Thomas
6ceed517de
end filename
2021-05-29 12:36:45 -04:00
Brandon Thomas
f886cecf8d
it works?? I need to save the results somewhere
2021-05-29 12:20:46 -04:00
Brandon Thomas
62b6b70e36
download template media, start work on process script
2021-05-29 12:07:05 -04:00
Brandon Thomas
6d7f9ba872
inference job (doesn't work)
2021-05-29 02:41:57 -04:00
Brandon Thomas
765d4b597d
record w2l file size
2021-05-27 00:56:52 -04:00
Brandon Thomas
43736d01e2
get original file mimetype
2021-05-25 02:40:27 -04:00
Brandon Thomas
6afe1c168b
save duration millis
2021-05-25 01:50:02 -04:00
Brandon Thomas
aacc6e5fb9
Read file metadata and populate DB
2021-05-25 00:54:49 -04:00
Brandon Thomas
6a07f5f91d
Preprocess W2L uploads (but glitchy)
2021-05-24 19:22:40 -04:00
Brandon Thomas
066cd01dd9
remove unnecessary steps from preprocessing
2021-05-24 15:11:40 -04:00
Brandon Thomas
2727afd644
Getting ready to process W2L uploads...
2021-05-24 15:00:10 -04:00
Brandon Thomas
a8f969dc55
add READMEs for w2l
2021-05-24 03:19:33 -04:00
Brandon Thomas
effe301d23
Add adapted Wav2Lip from https://github.com/ml-applications/wav2lip-edge
2021-05-24 03:17:53 -04:00