deploy DeepFilterNet denoise app
Browse files- README.md +4 -4
- app.py +45 -59
- requirements.txt +1 -2
README.md
CHANGED
|
@@ -1,13 +1,13 @@
|
|
| 1 |
---
|
| 2 |
-
title:
|
| 3 |
-
emoji:
|
| 4 |
colorFrom: green
|
| 5 |
colorTo: blue
|
| 6 |
sdk: gradio
|
| 7 |
sdk_version: 6.19.0
|
| 8 |
-
python_version:
|
| 9 |
app_file: app.py
|
| 10 |
pinned: false
|
| 11 |
---
|
| 12 |
|
| 13 |
-
|
|
|
|
| 1 |
---
|
| 2 |
+
title: 音频降噪增强
|
| 3 |
+
emoji: 🎧
|
| 4 |
colorFrom: green
|
| 5 |
colorTo: blue
|
| 6 |
sdk: gradio
|
| 7 |
sdk_version: 6.19.0
|
| 8 |
+
python_version: 3.13
|
| 9 |
app_file: app.py
|
| 10 |
pinned: false
|
| 11 |
---
|
| 12 |
|
| 13 |
+
基于 DeepFilterNet2 的人声降噪增强:上传带噪录音,下载干净版 wav。
|
app.py
CHANGED
|
@@ -1,67 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
-
from playwright.sync_api import sync_playwright
|
| 3 |
-
from PIL import Image
|
| 4 |
-
import io
|
| 5 |
-
import subprocess
|
| 6 |
-
import sys
|
| 7 |
-
import os
|
| 8 |
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
|
|
|
|
|
|
|
|
|
| 17 |
|
| 18 |
-
|
| 19 |
-
|
| 20 |
|
| 21 |
-
# ==================== 主函数 ====================
|
| 22 |
-
def run_automation(url: str, action: str = "screenshot"):
|
| 23 |
-
try:
|
| 24 |
-
with sync_playwright() as p:
|
| 25 |
-
browser = p.chromium.launch(headless=True)
|
| 26 |
-
page = browser.new_page(viewport={"width": 1280, "height": 720})
|
| 27 |
-
page.goto(url, wait_until="networkidle", timeout=60000)
|
| 28 |
-
|
| 29 |
-
if action == "screenshot":
|
| 30 |
-
screenshot_bytes = page.screenshot(full_page=True)
|
| 31 |
-
browser.close()
|
| 32 |
-
image = Image.open(io.BytesIO(screenshot_bytes))
|
| 33 |
-
return "✅ 截图成功!", image
|
| 34 |
-
else:
|
| 35 |
-
title = page.title()
|
| 36 |
-
browser.close()
|
| 37 |
-
return f"📄 页面标题:{title}", None
|
| 38 |
-
|
| 39 |
-
except Exception as e:
|
| 40 |
-
return f"❌ 错误:{str(e)}", None
|
| 41 |
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
with gr.Row():
|
| 47 |
-
url_input = gr.Textbox(
|
| 48 |
-
label="网址",
|
| 49 |
-
placeholder="https://huggingface.co",
|
| 50 |
-
value="https://example.com",
|
| 51 |
-
scale=4
|
| 52 |
-
)
|
| 53 |
-
action = gr.Radio(["screenshot", "title"], label="操作", value="screenshot")
|
| 54 |
-
|
| 55 |
-
btn = gr.Button("🚀 执行", variant="primary")
|
| 56 |
-
|
| 57 |
-
with gr.Row():
|
| 58 |
-
output_text = gr.Textbox(label="状态")
|
| 59 |
-
output_image = gr.Image(label="截图结果")
|
| 60 |
|
| 61 |
-
|
| 62 |
-
|
| 63 |
-
|
| 64 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 65 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 66 |
|
| 67 |
-
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python3
|
| 2 |
+
"""音频降噪增强:DeepFilterNet2 人声增强。
|
| 3 |
+
|
| 4 |
+
上传带噪音频(口播、讲座录音等),输出降噪后的 wav。
|
| 5 |
+
模型在首次请求时加载并缓存,之后常驻内存。
|
| 6 |
+
"""
|
| 7 |
+
import threading
|
| 8 |
+
|
| 9 |
import gradio as gr
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 10 |
|
| 11 |
+
_lock = threading.Lock()
|
| 12 |
+
_model = None
|
| 13 |
+
_df_state = None
|
| 14 |
+
|
| 15 |
+
|
| 16 |
+
def get_model():
|
| 17 |
+
global _model, _df_state
|
| 18 |
+
if _model is None:
|
| 19 |
+
with _lock:
|
| 20 |
+
if _model is None:
|
| 21 |
+
from df.enhance import init_df
|
| 22 |
|
| 23 |
+
_model, _df_state, _ = init_df()
|
| 24 |
+
return _model, _df_state
|
| 25 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 26 |
|
| 27 |
+
def denoise(audio_path):
|
| 28 |
+
if not audio_path:
|
| 29 |
+
return None
|
| 30 |
+
from df.enhance import enhance, load_audio, save_audio
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
|
| 32 |
+
model, df_state = get_model()
|
| 33 |
+
audio, _ = load_audio(audio_path, sr=df_state.sr())
|
| 34 |
+
enhanced = enhance(model, df_state, audio)
|
| 35 |
+
out = "/tmp/enhanced.wav"
|
| 36 |
+
save_audio(out, enhanced, df_state.sr())
|
| 37 |
+
return out
|
| 38 |
+
|
| 39 |
+
|
| 40 |
+
with gr.Blocks(title="音频降噪增强") as demo:
|
| 41 |
+
gr.Markdown(
|
| 42 |
+
"# 🎧 音频降噪增强\n"
|
| 43 |
+
"基于 DeepFilterNet2 的人声降噪:上传带底噪、环境声的录音,输出干净版 wav(48kHz)。\n\n"
|
| 44 |
+
"适合口播、讲座这类以人声为主的素材。CPU 推理,长音频请按时长耐心等。"
|
| 45 |
)
|
| 46 |
+
inp = gr.Audio(label="原始音频", type="filepath")
|
| 47 |
+
btn = gr.Button("开始降噪", variant="primary")
|
| 48 |
+
out = gr.Audio(label="降噪后", type="filepath")
|
| 49 |
+
btn.click(denoise, inputs=inp, outputs=out)
|
| 50 |
+
|
| 51 |
|
| 52 |
+
if __name__ == "__main__":
|
| 53 |
+
demo.launch()
|
requirements.txt
CHANGED
|
@@ -1,2 +1 @@
|
|
| 1 |
-
|
| 2 |
-
playwright
|
|
|
|
| 1 |
+
deepfilternet==0.5.6
|
|
|