Spaces:
Paused
Paused
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,67 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
from tempfile import gettempdir
|
3 |
+
import gradio as gr
|
4 |
+
import subprocess
|
5 |
+
os.system("git clone https://github.com/doevent/FullSubNet-plus")
|
6 |
+
os.system("mv FullSubNet-plus/speech_enhance .")
|
7 |
+
os.system("mv FullSubNet-plus/config .")
|
8 |
+
os.system("gdown https://drive.google.com/uc?id=1UJSt1G0P_aXry-u79LLU_l9tCnNa2u7C -O best_model.tar")
|
9 |
+
from speech_enhance.tools.denoise_server import start
|
10 |
+
|
11 |
+
|
12 |
+
# If the file is too duration to inference
|
13 |
+
def duration(input_audio) -> int:
|
14 |
+
command = f"ffprobe -v error -show_entries format=duration -of default=noprint_wrappers=1:nokey=1 -i {input_audio}"
|
15 |
+
result = subprocess.run(command, shell=True, stdout=subprocess.PIPE)
|
16 |
+
data = result.stdout.decode('ascii').rstrip()
|
17 |
+
return int(float(data))
|
18 |
+
|
19 |
+
|
20 |
+
def inference(audio):
|
21 |
+
try:
|
22 |
+
if duration(audio) >= 150:
|
23 |
+
return "error.wav"
|
24 |
+
result = start(to_list_files=[audio])
|
25 |
+
return result
|
26 |
+
except Exception as e:
|
27 |
+
return "error.wav"
|
28 |
+
|
29 |
+
|
30 |
+
title = """<h1 id="title">DeNoise Speech Enhancement</h1>"""
|
31 |
+
description = """
|
32 |
+
This is an unofficial demo for FullSubNet-plus: DeNoise Speech Enhancement. To use it, simply upload your audio, or click one of the examples to load them. Read more at the links below.
|
33 |
+
Link to GitHub:
|
34 |
+
- [FullSubNet +](https://github.com/hit-thusz-RookieCJ/FullSubNet-plus)
|
35 |
+
"""
|
36 |
+
twitter_link = "[](https://twitter.com/DoEvent)"
|
37 |
+
css = '''
|
38 |
+
h1#title {
|
39 |
+
text-align: center;
|
40 |
+
}
|
41 |
+
'''
|
42 |
+
|
43 |
+
demo = gr.Blocks(css=css)
|
44 |
+
with demo:
|
45 |
+
gr.Markdown(title)
|
46 |
+
gr.Markdown(description)
|
47 |
+
gr.Markdown(twitter_link)
|
48 |
+
|
49 |
+
|
50 |
+
with gr.Tabs():
|
51 |
+
with gr.TabItem("Upload audio"):
|
52 |
+
iface = gr.Interface(
|
53 |
+
inference,
|
54 |
+
inputs = gr.Audio(source="upload", type="filepath"),
|
55 |
+
outputs = gr.Audio(type="file"),
|
56 |
+
examples=[["man.wav"], ["woman.wav"]])
|
57 |
+
with gr.TabItem("Record your voice"):
|
58 |
+
iface = gr.Interface(
|
59 |
+
inference,
|
60 |
+
inputs = gr.Audio(source="microphone", label="Record yourself reading something out loud", type="filepath"),
|
61 |
+
outputs = gr.Audio(type="file"),
|
62 |
+
)
|
63 |
+
|
64 |
+
gr.Markdown("<div><center><img src='https://visitor-badge.glitch.me/badge?page_id=max_skobeev_fullsubnet_plus_public' alt='visitor badge'></center></div>")
|
65 |
+
|
66 |
+
|
67 |
+
demo.launch(enable_queue=True, debug=True, show_error=True)
|