Update README.md
Browse files
README.md
CHANGED
|
@@ -6,22 +6,43 @@ base_model:
|
|
| 6 |
- IndexTeam/IndexTTS-2
|
| 7 |
---
|
| 8 |
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
守ってほしいです。
|
| 13 |
|
| 14 |
-
|
| 15 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 16 |
|
| 17 |
-
|
| 18 |
-
|
| 19 |
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
japanese-bpe.modelをbpe.modelにリネームしてcheckpoints以下のデフォルトの物と置き換える
|
| 23 |
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
|
|
|
|
|
| 6 |
- IndexTeam/IndexTTS-2
|
| 7 |
---
|
| 8 |
|
| 9 |
+
train, infer (webgui) code in this fork
|
| 10 |
+
vram use lower than original and this model only work this repo
|
| 11 |
+
https://github.com/q9uri/index-tts-ja
|
|
|
|
| 12 |
|
| 13 |
+
model pretrain use jvnv courpus,
|
| 14 |
+
cretaed by taakamichi shinosuke sensei and japanese voice actor!
|
| 15 |
+
|
| 16 |
+
and reason-speech-v2-denoized
|
| 17 |
+
|
| 18 |
+
original reazon-speech was created by reazon team,
|
| 19 |
+
source voice was japanese tv
|
| 20 |
+
wav file under licensed by 日本国著作権の例外項目
|
| 21 |
+
|
| 22 |
+
denoized by fishaudio. use uvr5
|
| 23 |
+
reupload hf by litagin02
|
| 24 |
+
|
| 25 |
+
anime-whisper-0.3 use create text transcript
|
| 26 |
+
kanji in suppres token,transcripts nearly kana only text.
|
| 27 |
+
|
| 28 |
+
model was trained sigle rtx 3060 (max 60% setting,power look like rtx a2000)
|
| 29 |
+
batch size 1, amp don't use (haha, i forgotten. recommend use amp)
|
| 30 |
+
|
| 31 |
+
gpu was i have'n ローカルllmに向き合う会 hackason. thx サルドラ (@saldra) ゆづき
|
| 32 |
+
|
| 33 |
+
may i support me?
|
| 34 |
+
buy gpu for me amazon.jp [shoplist](https://www.amazon.jp/hz/wishlist/ls/1P9G04X6Z55BK?ref_=wl_shar)
|
| 35 |
+
|
| 36 |
+
download custom pretrain models
|
| 37 |
+
https://huggingface.co/WariHima/index-tts-japanese-prosody
|
| 38 |
|
| 39 |
+
and other orginal index-tts2 weight file need.
|
|
|
|
| 40 |
|
| 41 |
+
infer need cuda 12.8 and vram 8gb
|
| 42 |
+
created voice length * 2 sec,
|
|
|
|
| 43 |
|
| 44 |
+
36000+6000x8x6cycle_steps.pth rename to gpt.pth, copy to ./checkpoints.
|
| 45 |
+
japanese-bpe.model to ./checkpoints, don't be rename.
|
| 46 |
+
|
| 47 |
+
run webui
|
| 48 |
+
python webui.py
|