|
import io |
|
|
|
import gradio as gr |
|
import librosa |
|
import numpy as np |
|
import soundfile |
|
from inference.infer_tool import Svc |
|
import logging |
|
from logmmse import logmmse |
|
|
|
logging.getLogger('numba').setLevel(logging.WARNING) |
|
|
|
model_name = "logs/32k/uma1.pth" |
|
config_name = "configs/uma1.json" |
|
model2_name="logs/32k/uma2.pth" |
|
config2_name = "configs/uma2.json" |
|
model3_name="logs/32k/uma3.pth" |
|
config3_name = "configs/uma3.json" |
|
sid_map = { |
|
"米浴":"rice", |
|
"东海帝皇":"teio", |
|
"爱慕织姬":"aimya", |
|
"成田大进":"taishin", |
|
"优秀素质":"nature", |
|
"待兼诗歌剧":"mati", |
|
"大拓太阳神":"sun", |
|
"目白善信":"pama", |
|
"第一红宝石":"ruby" |
|
} |
|
|
|
svc = Svc(model_name, config_name) |
|
def vc_fn(sid, vc_input3,vc_input4,vc_transform,sid3): |
|
if(vc_input3==None and vc_input4==None): |
|
return "请上传一段音频后再次尝试", None |
|
if sid3=='文件': |
|
sampling_rate, audio = vc_input3 |
|
else: |
|
sampling_rate, audio = vc_input4 |
|
|
|
duration = audio.shape[0] / sampling_rate |
|
audio = (audio / np.iinfo(audio.dtype).max).astype(np.float32) |
|
if len(audio.shape) > 1: |
|
audio = librosa.to_mono(audio.transpose(1, 0)) |
|
if sampling_rate != 32000: |
|
audio = librosa.resample(audio, orig_sr=sampling_rate, target_sr=32000) |
|
audio = logmmse(audio, 32000) |
|
print(audio.shape) |
|
out_wav_path = io.BytesIO() |
|
soundfile.write(out_wav_path, audio, 32000, format="wav") |
|
out_wav_path.seek(0) |
|
sid = sid_map[sid] |
|
print(sid) |
|
|
|
if sid in ["rice","teio","aimya"]: |
|
svc = Svc(model_name, config_name) |
|
if sid in ["taishin","nature","mati"]: |
|
svc = Svc(model2_name, config2_name) |
|
if sid in ["sun","ruby","pama"]: |
|
svc = Svc(model3_name, config3_name) |
|
out_audio, _out_sr = svc.infer(sid, vc_transform, out_wav_path) |
|
|
|
_audio = out_audio.cpu().numpy() |
|
return sid3, (32000, _audio) |
|
|
|
|
|
app = gr.Blocks() |
|
with app: |
|
with gr.Tabs(): |
|
with gr.TabItem("Basic"): |
|
gr.Markdown(value=""" |
|
# 前言 |
|
本demo基于[sovits 3.0 32khz版本](https://github.com/innnky/so-vits-svc)训练的，并改写于 `https://huggingface.co/spaces/innnky/nyaru-svc-3.0`， |
|
https://huggingface.co/spaces/yukie/yukie-sovits3等 |
|
特别感谢innnky佬与yukie佬 |
|
加载赛马娘语音，自用。 |
|
""") |
|
sid = gr.Dropdown(label="音色", choices=[ |
|
"东海帝皇", "米浴","爱慕织姬", "成田大进","优秀素质","待兼诗歌剧","大拓太阳神","目白善信","第一红宝石"], value="东海帝皇") |
|
sid2 = gr.Dropdown(label="上传方式", choices=[ |
|
"文件", "录音"], value="文件") |
|
vc_input3 = gr.Audio(label="上传音频") |
|
vc_input4 = gr.Audio(source="microphone") |
|
vc_transform = gr.Number( |
|
label="变调（整数，可以正负，半音数量，升高八度就是12）", value=0) |
|
vc_submit = gr.Button("转换", variant="primary") |
|
vc_output1 = gr.Textbox(label="Output Message") |
|
vc_output2 = gr.Audio(label="Output Audio") |
|
|
|
gr.Markdown(value=""" |
|
## 注意 |
|
不要使用太长的语音 |
|
""") |
|
vc_submit.click(vc_fn, [sid,vc_input3,vc_input4, vc_transform,sid2], [ |
|
vc_output1, vc_output2]) |
|
|
|
app.launch() |
|
|
