add initial README and models
Browse files- README.md +47 -3
- brickwall-e32.ckpt +3 -0
- brickwall_e96_a168096.pth +3 -0
README.md
CHANGED
|
@@ -1,3 +1,47 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: mit
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: mit
|
| 3 |
+
language:
|
| 4 |
+
- ja
|
| 5 |
+
pipeline_tag: text-to-speech
|
| 6 |
+
---
|
| 7 |
+
Following this guide with exceptions
|
| 8 |
+
https://rentry.org/GPT-SoVITS-guide
|
| 9 |
+
|
| 10 |
+
I used the latest git pull from
|
| 11 |
+
https://github.com/RVC-Boss/GPT-SoVITS/
|
| 12 |
+
|
| 13 |
+
I needed to put:
|
| 14 |
+
|
| 15 |
+
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/path/to/cudann/lib/
|
| 16 |
+
|
| 17 |
+
in my shell.
|
| 18 |
+
|
| 19 |
+
|
| 20 |
+
put the pth file in your SoVITS_weights_v2 folder and the ckpt in GPT_weights_v2
|
| 21 |
+
|
| 22 |
+
make both Language for Reference audio and Inference text language "Japanese", set slicing to "Slice by every punct".
|
| 23 |
+
|
| 24 |
+
the underlying model was trained on audio that was brickwall dynamic range compressed, like that commonly found in visual novels or video games.
|
| 25 |
+
|
| 26 |
+
you should be able to give it a CLEAN AND NOISE/MUSIC/STATIC free FEMALE Japanese voice clip from 3-10 seconds, give it a 100% ACCURATE transcription and get ok results out the other side.
|
| 27 |
+
|
| 28 |
+
I have found that results can be improved applying post-generation noise reduction and some treble boosting EQ processing. Audacity works well enough for this, since there isn't anything in the official Gradio interface.
|
| 29 |
+
|
| 30 |
+
Feel free to keep everything else at the deaults
|
| 31 |
+
|
| 32 |
+
If you want to start the inference engine auomatically, you can use do something like
|
| 33 |
+
|
| 34 |
+
python3 /path/to/GPT_SoVITS/inference_webui.py "Auto"
|
| 35 |
+
|
| 36 |
+
If you isolate it ala https://rentry.org/IsolatedLinuxWebService and put nginx in front of it with an ssl cert, you need something like this in the location block:
|
| 37 |
+
|
| 38 |
+
proxy_pass http://127.0.0.1:9872/;
|
| 39 |
+
proxy_buffering off;
|
| 40 |
+
proxy_redirect off;
|
| 41 |
+
proxy_http_version 1.1;
|
| 42 |
+
proxy_set_header Upgrade $http_upgrade;
|
| 43 |
+
proxy_set_header Connection "upgrade";
|
| 44 |
+
proxy_set_header Host $host;
|
| 45 |
+
client_max_body_size 500M;
|
| 46 |
+
proxy_set_header X-Forwarded-Proto $scheme;
|
| 47 |
+
add_header 'Content-Security-Policy' 'upgrade-insecure-requests';
|
brickwall-e32.ckpt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ab384212e78adbefb59aca1cc99acc910f896eb43e5c3b6a031ccf5f67576631
|
| 3 |
+
size 155312957
|
brickwall_e96_a168096.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:19bb170ff1e2cf75c7be5235ee9b68210f7f48dd9d627267f96abb3068d89950
|
| 3 |
+
size 85007202
|