James Betker
550874cbec
Update broken train_empire voice
2022-05-19 10:26:46 -06:00
James Betker
8139afd0e5
Remove CVVP
...
After training a similar model for a different purpose, I realized that
this model is faulty: the contrastive loss it uses only pays attention
to high-frequency details which do not contribute meaningfully to
output quality. I validated this by comparing a no-CVVP output with
a baseline using tts-scores and found no differences.
2022-05-17 12:21:25 -06:00
James Betker
aef86d21bf
Add a way to get deterministic behavior from tortoise and add debug states for reporting
2022-05-17 12:11:18 -06:00
James Betker
9eac62598a
Merge remote-tracking branch 'origin/main'
2022-05-17 11:22:40 -06:00
James Betker
24612f81c2
Add chapter 1 of GoT for read.py demos
2022-05-17 11:21:57 -06:00
James Betker
160963b105
Add conditioning latent example
2022-05-17 11:21:37 -06:00
Danila Berezin
dc3d7b1667
Fix bug in load_voices in audio.py
...
The read.py script did not work with pth latents, so I fix bug in audio.py. It seems that in the elif statement, instead of voice, voices should be clip, clips. And torch stack doesn't work with tuples, so I had to split this operation.
2022-05-17 18:34:54 +03:00
James Betker
11e80b0dae
Merge pull request #42 from jnordberg/main
...
Improve sentence splitting
2022-05-14 08:52:46 -06:00
James Betker
50690e4465
Automatically pick batch size based on available GPU memory
2022-05-13 10:30:02 -06:00
James Betker
cb7adf16af
Remove samples_generator
...
Was useful but the page is more detailed now.
2022-05-13 10:28:16 -06:00
Johan Nordberg
5197904660
Improve sentence splitting
2022-05-13 11:02:17 +00:00
James Betker
1a4f0fa350
update model paths (including clvp2!)
2022-05-12 20:18:11 -06:00
James Betker
75b0e03ab3
Add error message
2022-05-12 20:15:40 -06:00
James Betker
ec16c0208c
add eval script for testing
2022-05-12 20:15:22 -06:00
James Betker
7d5e7dbba8
CLVP2!
2022-05-12 13:23:03 -06:00
James Betker
fda5130819
Add support for multiple output candidates in do_tts.
2022-05-12 11:25:35 -06:00
James Betker
0005d02940
read.py: allow user-specified splits
2022-05-12 11:24:55 -06:00
Mark Baushenko
cbccc5e953
Optimizing graphics card memory
...
During inference it does not store gradients, which take up most of the video memory
2022-05-11 16:35:11 +03:00
James Betker
317d55c252
re-enable redaction
2022-05-06 09:36:42 -06:00
James Betker
8672075914
temporarily disable redaction
2022-05-06 09:06:20 -06:00
James Betker
d03b8f3f62
Update rainbow
2022-05-06 08:55:08 -06:00
James Betker
ffd0238a16
v2.2
2022-05-06 00:11:10 -06:00
James Betker
b327be56c6
Update readme with prompt engineering
2022-05-03 21:32:06 -06:00
James Betker
ddb19f6b0f
Enable redaction by default
2022-05-03 21:21:52 -06:00
James Betker
c1d004aeb0
change quality presets
2022-05-03 21:01:26 -06:00
James Betker
ee6f9b15ce
Use librosa for loading mp3s
2022-05-03 20:44:31 -06:00
James Betker
29b2f36f55
Remove entmax dep
2022-05-02 21:43:14 -06:00
James Betker
14617f8963
Fix default output path
2022-05-02 21:37:39 -06:00
James Betker
8c7f709c12
k I think this works..
2022-05-02 21:31:31 -06:00
James Betker
9acce239d3
fix paths
2022-05-02 20:56:28 -06:00
James Betker
a4cda68ddf
getting ready for 2.1 release
2022-05-02 20:20:50 -06:00
James Betker
f499d66493
misc fixes
2022-05-02 18:00:57 -06:00
James Betker
2888ae0337
Fix bug with k>1
2022-05-02 18:00:22 -06:00
James Betker
cdf44d7506
more fixes
2022-05-02 16:44:47 -06:00
James Betker
5a95c34a01
fix warning
2022-05-02 16:36:02 -06:00
James Betker
39ec1b0db5
Support totally random voices (and make fixes to previous changes)
2022-05-02 15:40:03 -06:00
James Betker
7d5bb8959a
move data
2022-05-02 15:39:34 -06:00
James Betker
9007955d88
Add redaction support
2022-05-02 14:57:29 -06:00
James Betker
cd2d4229bf
Better error messages when inputs are out of bounds.
2022-05-01 17:39:36 -06:00
James Betker
b1fc2b13c9
add support for specifying the model_dir
2022-05-01 17:29:25 -06:00
James Betker
0ffc191408
Add support for extracting and feeding conditioning latents directly into the model
...
- Adds a new script and API endpoints for doing this
- Reworks autoregressive and diffusion models so that the conditioning is computed separately (which will actually provide a mild performance boost)
- Updates README
This is untested. Need to do the following manual tests (and someday write unit tests for this behemoth before
it becomes a problem..)
1) Does get_conditioning_latents.py work?
2) Can I feed those latents back into the model by creating a new voice?
3) Can I still mix and match voices (both with conditioning latents and normal voices) with read.py?
2022-05-01 17:25:18 -06:00
James Betker
a8264f5cef
more cleanup
2022-05-01 16:28:39 -06:00
James Betker
f7c8decfdb
Move everything into the tortoise/ subdirectory
...
For eventual packaging.
2022-05-01 16:24:24 -06:00
