Spaces:
Running
Running
import gradio as gr | |
import os | |
import pytz | |
from datetime import datetime | |
import pandas as pd | |
import plotly.express as px | |
from svc_inference import main | |
# データの作成 | |
data = { | |
'id': list(range(1, 57)), # 1から56までの数字 | |
'x': [ | |
28, 25, 5, 12, 8, 2, 0, -20, -15, -12, -20, 8, -30, 25, 0, 0, 2, -25, | |
-25, 20, 15, -2, 0, 15, -30, 15, 8, 28, -10, -22, 20, 20, 8, 20, 0, | |
0, -8, -10, -32, 0, 0, -8, 2, -25, -32, -20, -18, -5, 15, -22, -25, | |
-28, -30, 10, 25, 28 | |
], | |
'y': [ | |
0, -5, -15, -20, -18, -3, 8, 8, 12, 10, 10, -20, 6, -3, 12, -15, 12, | |
17, 10, -8, -15, -22, 8, 15, 10, -15, -18, -10, 8, 5, -10, -8, -25, | |
-5, -12, 12, 15, 6, 17, -12, -8, -8, 15, 17, 25, 4, 4, 0, 0, -20, | |
12, 12, 15, -19, 0, 0 | |
] | |
} | |
df = pd.DataFrame(data) | |
def create_plot(): | |
fig = px.scatter(df, x='x', y='y', text='id', | |
title='Interactive Numeric Scatter Plot') | |
# マーカーのスタイルを設定(紫系の色に設定) | |
fig.update_traces( | |
marker=dict( | |
size=10, | |
color='#663399', # 紫色 | |
line=dict(color='#4B0082', width=1) # より暗い紫の境界線 | |
), | |
textposition='top center' | |
) | |
# レイアウトの設定 | |
fig.update_layout( | |
height=600, | |
width=800, | |
clickmode='event+select', | |
plot_bgcolor='#eeeeee', | |
paper_bgcolor='white', | |
xaxis=dict( | |
showgrid=True, | |
zeroline=True, | |
range=[-35, 35] # x軸の範囲を設定 | |
), | |
yaxis=dict( | |
showgrid=True, | |
zeroline=True, | |
range=[-30, 30] # y軸の範囲を設定 | |
) | |
) | |
return fig | |
def run_main(audio_file, shift, speaker_id): | |
# 固定の引数を設定 | |
class Args: | |
pass | |
args = Args() | |
args.config = "configs/base.yaml" | |
args.model = "./vits_pretrain/sovits5.0.pretrain.pth" | |
speaker_str = f"{speaker_id:04d}" | |
args.spk = f"./configs/singers/singer{speaker_str}.npy" | |
args.wave = audio_file | |
print(audio_file) | |
args.shift = shift | |
# オプショナルパラメータのデフォルト値設定 | |
args.ppg = None | |
args.vec = None | |
args.pit = None | |
args.enable_retrieval = False | |
args.retrieval_index_prefix = "" | |
args.retrieval_ratio = 0.5 | |
args.n_retrieval_vectors = 3 | |
args.hubert_index_path = None | |
args.whisper_index_path = None | |
args.debug = False | |
# 現在時刻を取得してファイル名を生成 | |
jst = pytz.timezone('Asia/Tokyo') | |
now = datetime.now(jst) | |
date_str = now.strftime("%y%m%d") | |
time_str = now.strftime("%H%M%S") | |
output_filename = f"svc_id{speaker_str}_pitch{shift}_speed0_{date_str}_{time_str}.wav" | |
try: | |
main(args) | |
os.rename("svc_out.wav", output_filename) # ファイル名を変更 | |
return output_filename # 新しいファイル名を返す | |
except Exception as e: | |
return None | |
# Gradio インターフェースの作成 | |
with gr.Blocks() as demo: | |
gr.Markdown("# SVC 音声変換システム") | |
with gr.Row(): | |
with gr.Column(scale=1.15): | |
plot = gr.Plot(value=create_plot()) | |
with gr.Column(scale=1): | |
# 入力音声のアップロード | |
input_audio = gr.Audio( | |
label="変換したい音声をアップロード", | |
type="filepath" # ファイルパスとして取得 | |
) | |
# Speaker ID の選択 | |
speaker_id = gr.Number( | |
label="Speaker ID (1-56)", | |
value=1, | |
minimum=1, | |
maximum=56, | |
step=1 | |
) | |
# Pitch シフトのスライダー | |
shift = gr.Slider( | |
minimum=-12, | |
maximum=12, | |
value=0, | |
step=1, | |
label="Pitch Shift(-12から+12の半音)" | |
) | |
# ボタン | |
run_btn = gr.Button(value="音声変換を実行", variant="primary", size="lg") | |
# 出力表示用 | |
output_audio = gr.Audio(label="変換後の音声") | |
run_btn.click( | |
fn=run_main, | |
inputs=[input_audio, shift, speaker_id], | |
outputs=[output_audio] | |
) | |
# アプリケーションの起動 | |
if __name__ == "__main__": | |
demo.launch() | |