Spaces:
Running
Running
feat: 重构云端 Web UI,实现上传处理下载流程
Browse files- app.py +3 -3
- docs/流程文档_AI用.md +24 -8
- main_local.py +13 -0
- src/gui_cloud.py +685 -0
app.py
CHANGED
|
@@ -103,10 +103,10 @@ def main():
|
|
| 103 |
"""主入口"""
|
| 104 |
setup_environment()
|
| 105 |
|
| 106 |
-
# 导入并启动 GUI
|
| 107 |
-
from src.
|
| 108 |
|
| 109 |
-
app =
|
| 110 |
|
| 111 |
# 云端配置
|
| 112 |
app.launch(
|
|
|
|
| 103 |
"""主入口"""
|
| 104 |
setup_environment()
|
| 105 |
|
| 106 |
+
# 导入并启动云端 GUI
|
| 107 |
+
from src.gui_cloud import create_cloud_ui
|
| 108 |
|
| 109 |
+
app = create_cloud_ui()
|
| 110 |
|
| 111 |
# 云端配置
|
| 112 |
app.launch(
|
docs/流程文档_AI用.md
CHANGED
|
@@ -37,7 +37,8 @@
|
|
| 37 |
|
| 38 |
```
|
| 39 |
项目根目录/
|
| 40 |
-
├── main.py # 程序入口
|
|
|
|
| 41 |
├── config.json # 全局配置文件
|
| 42 |
├── bank/ # 音源库目录
|
| 43 |
│ └── [音源名称]/
|
|
@@ -220,7 +221,7 @@ MFA 支持两种运行模式:
|
|
| 220 |
|
| 221 |
## 使用流程
|
| 222 |
|
| 223 |
-
### 方式一: 本地 Web UI
|
| 224 |
|
| 225 |
1. 运行 `python main.py` 启动 Web UI
|
| 226 |
2. 浏览器自动打开 http://127.0.0.1:7860
|
|
@@ -239,9 +240,15 @@ MFA 支持两种运行模式:
|
|
| 239 |
- 配置导出选项并执行
|
| 240 |
- 点击下载按钮获取结果
|
| 241 |
|
| 242 |
-
> 注: 旧版 CustomTkinter 桌面 GUI 已移至 `src/gui_old.py`
|
| 243 |
|
| 244 |
-
### 方式二:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 245 |
|
| 246 |
1. 使用 `app.py` 作为入口文件
|
| 247 |
2. 云端环境自动安装 MFA 和下载模型
|
|
@@ -251,7 +258,7 @@ MFA 支持两种运行模式:
|
|
| 251 |
- Hugging Face Spaces (Gradio SDK)
|
| 252 |
- 魔塔社区 ModelScope (推荐,国内访问快)
|
| 253 |
|
| 254 |
-
### 方式
|
| 255 |
|
| 256 |
```python
|
| 257 |
from src.pipeline import PipelineConfig, VoiceBankPipeline
|
|
@@ -295,16 +302,25 @@ MFA 环境:
|
|
| 295 |
|
| 296 |
```
|
| 297 |
项目根目录/
|
| 298 |
-
├── app.py # 云端入口 (
|
| 299 |
-
├── main.py # 本地入口
|
| 300 |
├── requirements.txt
|
| 301 |
├── src/
|
| 302 |
-
│ ├── gui.py #
|
|
|
|
|
|
|
| 303 |
│ ├── mfa_runner.py # 跨平台 MFA 调用
|
| 304 |
│ └── ...
|
| 305 |
└── ...
|
| 306 |
```
|
| 307 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 308 |
### 平台差异
|
| 309 |
|
| 310 |
| 功能 | 本地 (Windows) | 云端 (Linux) |
|
|
|
|
| 37 |
|
| 38 |
```
|
| 39 |
项目根目录/
|
| 40 |
+
├── main.py # 程序入口 (Web UI)
|
| 41 |
+
├── main_local.py # 本地桌面入口 (CustomTkinter GUI)
|
| 42 |
├── config.json # 全局配置文件
|
| 43 |
├── bank/ # 音源库目录
|
| 44 |
│ └── [音源名称]/
|
|
|
|
| 221 |
|
| 222 |
## 使用流程
|
| 223 |
|
| 224 |
+
### 方式一: 本地 Web UI (Gradio)
|
| 225 |
|
| 226 |
1. 运行 `python main.py` 启动 Web UI
|
| 227 |
2. 浏览器自动打开 http://127.0.0.1:7860
|
|
|
|
| 240 |
- 配置导出选项并执行
|
| 241 |
- 点击下载按钮获取结果
|
| 242 |
|
| 243 |
+
> 注: 旧版 CustomTkinter 桌面 GUI 已移至 `src/gui_old.py`,可通过 `python main_local.py` 启动
|
| 244 |
|
| 245 |
+
### 方式二: 本地桌面 GUI (CustomTkinter)
|
| 246 |
+
|
| 247 |
+
1. 运行 `python main_local.py` 启动桌面应用
|
| 248 |
+
2. 使用原生窗口界面操作,功能与 Web UI 相同
|
| 249 |
+
3. 适合不需要浏览器的本地独立运行场景
|
| 250 |
+
|
| 251 |
+
### 方式三: 云端部署 (HF Spaces / 魔塔社区)
|
| 252 |
|
| 253 |
1. 使用 `app.py` 作为入口文件
|
| 254 |
2. 云端环境自动安装 MFA 和下载模型
|
|
|
|
| 258 |
- Hugging Face Spaces (Gradio SDK)
|
| 259 |
- 魔塔社区 ModelScope (推荐,国内访问快)
|
| 260 |
|
| 261 |
+
### 方式四: 命令行/脚本
|
| 262 |
|
| 263 |
```python
|
| 264 |
from src.pipeline import PipelineConfig, VoiceBankPipeline
|
|
|
|
| 302 |
|
| 303 |
```
|
| 304 |
项目根目录/
|
| 305 |
+
├── app.py # 云端入口 (使用 gui_cloud.py)
|
| 306 |
+
├── main.py # 本地入口 (使用 gui.py)
|
| 307 |
├── requirements.txt
|
| 308 |
├── src/
|
| 309 |
+
│ ├── gui.py # 本地 Web UI (完整功能)
|
| 310 |
+
│ ├── gui_cloud.py # 云端 Web UI (上传→处理→下载)
|
| 311 |
+
│ ├── gui_old.py # 旧版桌面 GUI (CustomTkinter)
|
| 312 |
│ ├── mfa_runner.py # 跨平台 MFA 调用
|
| 313 |
│ └── ...
|
| 314 |
└── ...
|
| 315 |
```
|
| 316 |
|
| 317 |
+
### 云端 GUI 特点 (gui_cloud.py)
|
| 318 |
+
|
| 319 |
+
- **制作音源**: 上传音频文件 → VAD切片 + Whisper转录 + MFA对齐 → 下载音源包
|
| 320 |
+
- **导出音源**: 上传音源包 → 选择导出插件 → 下载导出结果
|
| 321 |
+
- 使用临时工作空间,处理完成后自动清理
|
| 322 |
+
- 无需本地持久化存储
|
| 323 |
+
|
| 324 |
### 平台差异
|
| 325 |
|
| 326 |
| 功能 | 本地 (Windows) | 云端 (Linux) |
|
main_local.py
ADDED
|
@@ -0,0 +1,13 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# -*- coding: utf-8 -*-
|
| 2 |
+
"""
|
| 3 |
+
人力V助手 (JinrikiHelper) - 本地版入口 (CustomTkinter GUI)
|
| 4 |
+
作者:TNOT
|
| 5 |
+
开源协议:MIT
|
| 6 |
+
|
| 7 |
+
使用旧版 CustomTkinter 界面,适合本地独立运行
|
| 8 |
+
"""
|
| 9 |
+
|
| 10 |
+
from src.gui_old import main
|
| 11 |
+
|
| 12 |
+
if __name__ == "__main__":
|
| 13 |
+
main()
|
src/gui_cloud.py
ADDED
|
@@ -0,0 +1,685 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# -*- coding: utf-8 -*-
|
| 2 |
+
"""
|
| 3 |
+
人力V助手 (JinrikiHelper) 云端 Web UI
|
| 4 |
+
基于 Gradio 6.2.0 构建
|
| 5 |
+
专为云端部署优化:上传 → 处理 → 下载
|
| 6 |
+
|
| 7 |
+
作者:TNOT
|
| 8 |
+
"""
|
| 9 |
+
|
| 10 |
+
import gradio as gr
|
| 11 |
+
import logging
|
| 12 |
+
import os
|
| 13 |
+
import sys
|
| 14 |
+
import json
|
| 15 |
+
import tempfile
|
| 16 |
+
import zipfile
|
| 17 |
+
import shutil
|
| 18 |
+
import uuid
|
| 19 |
+
from pathlib import Path
|
| 20 |
+
from typing import Optional, List, Dict, Tuple
|
| 21 |
+
|
| 22 |
+
# 配置日志
|
| 23 |
+
logging.basicConfig(
|
| 24 |
+
level=logging.INFO,
|
| 25 |
+
format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
|
| 26 |
+
handlers=[logging.StreamHandler(sys.stdout)]
|
| 27 |
+
)
|
| 28 |
+
logger = logging.getLogger(__name__)
|
| 29 |
+
|
| 30 |
+
# 项目根目录
|
| 31 |
+
BASE_DIR = Path(__file__).parent.parent.absolute()
|
| 32 |
+
|
| 33 |
+
|
| 34 |
+
class CloudConfig:
|
| 35 |
+
"""云端配置"""
|
| 36 |
+
|
| 37 |
+
# 临时工作目录
|
| 38 |
+
TEMP_BASE = tempfile.gettempdir()
|
| 39 |
+
|
| 40 |
+
# 模型目录(云端使用项目内目录)
|
| 41 |
+
MODELS_DIR = str(BASE_DIR / "models")
|
| 42 |
+
MFA_DIR = str(BASE_DIR / "models" / "mfa")
|
| 43 |
+
|
| 44 |
+
# 支持的音频格式
|
| 45 |
+
AUDIO_EXTENSIONS = ('.wav', '.mp3', '.flac', '.ogg', '.m4a')
|
| 46 |
+
|
| 47 |
+
# Whisper 模型选项
|
| 48 |
+
WHISPER_MODELS = {
|
| 49 |
+
"whisper-small": "openai/whisper-small",
|
| 50 |
+
"whisper-medium": "openai/whisper-medium"
|
| 51 |
+
}
|
| 52 |
+
|
| 53 |
+
# 语言选项
|
| 54 |
+
LANGUAGES = ["chinese", "japanese"]
|
| 55 |
+
|
| 56 |
+
|
| 57 |
+
def create_temp_workspace() -> str:
|
| 58 |
+
"""创建临时工作空间"""
|
| 59 |
+
workspace_id = str(uuid.uuid4())[:8]
|
| 60 |
+
workspace = os.path.join(CloudConfig.TEMP_BASE, f"jinriki_{workspace_id}")
|
| 61 |
+
os.makedirs(workspace, exist_ok=True)
|
| 62 |
+
return workspace
|
| 63 |
+
|
| 64 |
+
|
| 65 |
+
def cleanup_workspace(workspace: str):
|
| 66 |
+
"""清理工作空间"""
|
| 67 |
+
if workspace and os.path.exists(workspace):
|
| 68 |
+
try:
|
| 69 |
+
shutil.rmtree(workspace)
|
| 70 |
+
logger.info(f"已清理工作空间: {workspace}")
|
| 71 |
+
except Exception as e:
|
| 72 |
+
logger.warning(f"清理工作空间失败: {e}")
|
| 73 |
+
|
| 74 |
+
|
| 75 |
+
def create_zip(source_dir: str, zip_name: str) -> Optional[str]:
|
| 76 |
+
"""打包目录为 zip"""
|
| 77 |
+
if not os.path.isdir(source_dir):
|
| 78 |
+
return None
|
| 79 |
+
try:
|
| 80 |
+
zip_path = os.path.join(CloudConfig.TEMP_BASE, f"{zip_name}.zip")
|
| 81 |
+
with zipfile.ZipFile(zip_path, 'w', zipfile.ZIP_DEFLATED) as zf:
|
| 82 |
+
for root, dirs, files in os.walk(source_dir):
|
| 83 |
+
for file in files:
|
| 84 |
+
file_path = os.path.join(root, file)
|
| 85 |
+
arcname = os.path.relpath(file_path, source_dir)
|
| 86 |
+
zf.write(file_path, arcname)
|
| 87 |
+
return zip_path
|
| 88 |
+
except Exception as e:
|
| 89 |
+
logger.error(f"打包失败: {e}")
|
| 90 |
+
return None
|
| 91 |
+
|
| 92 |
+
|
| 93 |
+
def extract_zip(zip_path: str, target_dir: str) -> Tuple[bool, str]:
|
| 94 |
+
"""解压 zip 文件"""
|
| 95 |
+
try:
|
| 96 |
+
with zipfile.ZipFile(zip_path, 'r') as zf:
|
| 97 |
+
zf.extractall(target_dir)
|
| 98 |
+
return True, "解压成功"
|
| 99 |
+
except Exception as e:
|
| 100 |
+
return False, f"解压失败: {e}"
|
| 101 |
+
|
| 102 |
+
|
| 103 |
+
def scan_mfa_models() -> Dict[str, List[str]]:
|
| 104 |
+
"""扫描 MFA 模型"""
|
| 105 |
+
result = {"acoustic": [], "dictionary": []}
|
| 106 |
+
if not os.path.exists(CloudConfig.MFA_DIR):
|
| 107 |
+
return result
|
| 108 |
+
for f in os.listdir(CloudConfig.MFA_DIR):
|
| 109 |
+
if f.endswith('.zip'):
|
| 110 |
+
result["acoustic"].append(f)
|
| 111 |
+
elif f.endswith('.dict') or f.endswith('.txt'):
|
| 112 |
+
result["dictionary"].append(f)
|
| 113 |
+
return result
|
| 114 |
+
|
| 115 |
+
|
| 116 |
+
def check_mfa_available() -> bool:
|
| 117 |
+
"""检查 MFA 是否可用"""
|
| 118 |
+
from src.mfa_runner import check_mfa_available as _check
|
| 119 |
+
return _check()
|
| 120 |
+
|
| 121 |
+
|
| 122 |
+
# ==================== 制作音源功能 ====================
|
| 123 |
+
|
| 124 |
+
def validate_audio_upload(files) -> Tuple[bool, str, List[str]]:
|
| 125 |
+
"""
|
| 126 |
+
验证上传的音频文件
|
| 127 |
+
|
| 128 |
+
返回: (是否有效, 消息, 文件路径列表)
|
| 129 |
+
"""
|
| 130 |
+
if not files:
|
| 131 |
+
return False, "请上传音频文件", []
|
| 132 |
+
|
| 133 |
+
valid_files = []
|
| 134 |
+
for f in files:
|
| 135 |
+
if hasattr(f, 'name'):
|
| 136 |
+
path = f.name
|
| 137 |
+
else:
|
| 138 |
+
path = str(f)
|
| 139 |
+
|
| 140 |
+
if path.lower().endswith(CloudConfig.AUDIO_EXTENSIONS):
|
| 141 |
+
valid_files.append(path)
|
| 142 |
+
|
| 143 |
+
if not valid_files:
|
| 144 |
+
return False, f"未找到有效音频文件,支持格式: {', '.join(CloudConfig.AUDIO_EXTENSIONS)}", []
|
| 145 |
+
|
| 146 |
+
return True, f"找到 {len(valid_files)} 个音频文件", valid_files
|
| 147 |
+
|
| 148 |
+
|
| 149 |
+
def process_make_voicebank(
|
| 150 |
+
audio_files,
|
| 151 |
+
source_name: str,
|
| 152 |
+
language: str,
|
| 153 |
+
whisper_model: str,
|
| 154 |
+
progress=gr.Progress()
|
| 155 |
+
) -> Tuple[str, str, Optional[str]]:
|
| 156 |
+
"""
|
| 157 |
+
制作音源:上传音频 → VAD切片 → Whisper转录 → MFA对齐 → 打包下载
|
| 158 |
+
|
| 159 |
+
返回: (状态, 日志, 下载文件路径)
|
| 160 |
+
"""
|
| 161 |
+
from src.pipeline import PipelineConfig, VoiceBankPipeline
|
| 162 |
+
|
| 163 |
+
logs = []
|
| 164 |
+
def log(msg):
|
| 165 |
+
logs.append(msg)
|
| 166 |
+
logger.info(msg)
|
| 167 |
+
|
| 168 |
+
# 验证输入
|
| 169 |
+
if not source_name or not source_name.strip():
|
| 170 |
+
return "❌ 请输入音源名称", "", None
|
| 171 |
+
|
| 172 |
+
source_name = source_name.strip()
|
| 173 |
+
|
| 174 |
+
valid, msg, file_paths = validate_audio_upload(audio_files)
|
| 175 |
+
if not valid:
|
| 176 |
+
return f"❌ {msg}", "", None
|
| 177 |
+
|
| 178 |
+
log(f"📁 {msg}")
|
| 179 |
+
|
| 180 |
+
# 创建临时工作空间
|
| 181 |
+
workspace = create_temp_workspace()
|
| 182 |
+
log(f"🔧 创建工作空间: {workspace}")
|
| 183 |
+
|
| 184 |
+
try:
|
| 185 |
+
# 准备输入目录
|
| 186 |
+
input_dir = os.path.join(workspace, "input")
|
| 187 |
+
bank_dir = os.path.join(workspace, "bank")
|
| 188 |
+
os.makedirs(input_dir, exist_ok=True)
|
| 189 |
+
os.makedirs(bank_dir, exist_ok=True)
|
| 190 |
+
|
| 191 |
+
# 复制音频文件到输入目录
|
| 192 |
+
progress(0.05, desc="复制音频文件...")
|
| 193 |
+
for src_path in file_paths:
|
| 194 |
+
dst_path = os.path.join(input_dir, os.path.basename(src_path))
|
| 195 |
+
shutil.copy2(src_path, dst_path)
|
| 196 |
+
log(f"📋 已复制 {len(file_paths)} 个文件到工作目录")
|
| 197 |
+
|
| 198 |
+
# 获取 MFA 模型路径
|
| 199 |
+
mfa_models = scan_mfa_models()
|
| 200 |
+
dict_path = None
|
| 201 |
+
acoustic_path = None
|
| 202 |
+
|
| 203 |
+
if mfa_models["dictionary"]:
|
| 204 |
+
# 根据语言选择字典
|
| 205 |
+
for d in mfa_models["dictionary"]:
|
| 206 |
+
if language == "japanese" and "japanese" in d.lower():
|
| 207 |
+
dict_path = os.path.join(CloudConfig.MFA_DIR, d)
|
| 208 |
+
break
|
| 209 |
+
elif language == "chinese" and "mandarin" in d.lower():
|
| 210 |
+
dict_path = os.path.join(CloudConfig.MFA_DIR, d)
|
| 211 |
+
break
|
| 212 |
+
if not dict_path:
|
| 213 |
+
dict_path = os.path.join(CloudConfig.MFA_DIR, mfa_models["dictionary"][0])
|
| 214 |
+
|
| 215 |
+
if mfa_models["acoustic"]:
|
| 216 |
+
for a in mfa_models["acoustic"]:
|
| 217 |
+
if language == "japanese" and "japanese" in a.lower():
|
| 218 |
+
acoustic_path = os.path.join(CloudConfig.MFA_DIR, a)
|
| 219 |
+
break
|
| 220 |
+
elif language == "chinese" and "mandarin" in a.lower():
|
| 221 |
+
acoustic_path = os.path.join(CloudConfig.MFA_DIR, a)
|
| 222 |
+
break
|
| 223 |
+
if not acoustic_path:
|
| 224 |
+
acoustic_path = os.path.join(CloudConfig.MFA_DIR, mfa_models["acoustic"][0])
|
| 225 |
+
|
| 226 |
+
# 配置流水线
|
| 227 |
+
whisper_model_name = CloudConfig.WHISPER_MODELS.get(whisper_model, "openai/whisper-small")
|
| 228 |
+
|
| 229 |
+
config = PipelineConfig(
|
| 230 |
+
source_name=source_name,
|
| 231 |
+
input_path=input_dir,
|
| 232 |
+
output_base_dir=bank_dir,
|
| 233 |
+
models_dir=CloudConfig.MODELS_DIR,
|
| 234 |
+
whisper_model=whisper_model_name,
|
| 235 |
+
mfa_dict_path=dict_path,
|
| 236 |
+
mfa_model_path=acoustic_path,
|
| 237 |
+
language=language
|
| 238 |
+
)
|
| 239 |
+
|
| 240 |
+
pipeline = VoiceBankPipeline(config, log)
|
| 241 |
+
|
| 242 |
+
# 步骤0: VAD切片 + Whisper转录
|
| 243 |
+
progress(0.1, desc="VAD切片 + Whisper转录...")
|
| 244 |
+
log("\n" + "=" * 50)
|
| 245 |
+
log("【步骤1】VAD切片 + Whisper转录")
|
| 246 |
+
success, msg, slices = pipeline.step0_preprocess()
|
| 247 |
+
if not success:
|
| 248 |
+
return f"❌ 预处理失败: {msg}", "\n".join(logs), None
|
| 249 |
+
log(f"✅ {msg}")
|
| 250 |
+
|
| 251 |
+
# 步骤1: MFA对齐
|
| 252 |
+
progress(0.6, desc="MFA语音对齐...")
|
| 253 |
+
log("\n" + "=" * 50)
|
| 254 |
+
log("【步骤2】MFA语音对齐")
|
| 255 |
+
|
| 256 |
+
if check_mfa_available():
|
| 257 |
+
success, msg = pipeline.step1_mfa_align()
|
| 258 |
+
if not success:
|
| 259 |
+
log(f"⚠️ MFA对齐失败: {msg}")
|
| 260 |
+
log("继续导出(无TextGrid)...")
|
| 261 |
+
else:
|
| 262 |
+
log(f"✅ {msg}")
|
| 263 |
+
else:
|
| 264 |
+
log("⚠️ MFA不可用,跳过对齐步骤")
|
| 265 |
+
|
| 266 |
+
# 打包结果
|
| 267 |
+
progress(0.9, desc="打包结果...")
|
| 268 |
+
log("\n" + "=" * 50)
|
| 269 |
+
log("【打包结果】")
|
| 270 |
+
|
| 271 |
+
source_dir = os.path.join(bank_dir, source_name)
|
| 272 |
+
zip_name = f"{source_name}_音源数据"
|
| 273 |
+
zip_path = create_zip(source_dir, zip_name)
|
| 274 |
+
|
| 275 |
+
if zip_path:
|
| 276 |
+
log(f"📦 已打包: {os.path.basename(zip_path)}")
|
| 277 |
+
progress(1.0, desc="完成")
|
| 278 |
+
return "✅ 音源制作完成", "\n".join(logs), zip_path
|
| 279 |
+
else:
|
| 280 |
+
return "❌ 打包失败", "\n".join(logs), None
|
| 281 |
+
|
| 282 |
+
except Exception as e:
|
| 283 |
+
logger.error(f"制作音源失败: {e}", exc_info=True)
|
| 284 |
+
return f"❌ 处理失败: {e}", "\n".join(logs), None
|
| 285 |
+
|
| 286 |
+
finally:
|
| 287 |
+
# 清理工作空间(保留zip文件)
|
| 288 |
+
cleanup_workspace(workspace)
|
| 289 |
+
|
| 290 |
+
|
| 291 |
+
# ==================== 导出音源功能 ====================
|
| 292 |
+
|
| 293 |
+
def validate_voicebank_zip(zip_file) -> Tuple[bool, str, Optional[str]]:
|
| 294 |
+
"""
|
| 295 |
+
验证上传的音源压缩包
|
| 296 |
+
|
| 297 |
+
返回: (是否有效, 消息, 音源名称)
|
| 298 |
+
"""
|
| 299 |
+
if not zip_file:
|
| 300 |
+
return False, "请上传音源压缩包", None
|
| 301 |
+
|
| 302 |
+
zip_path = zip_file.name if hasattr(zip_file, 'name') else str(zip_file)
|
| 303 |
+
|
| 304 |
+
if not zip_path.lower().endswith('.zip'):
|
| 305 |
+
return False, "请上传 .zip 格式的压缩包", None
|
| 306 |
+
|
| 307 |
+
# 检查压缩包内容
|
| 308 |
+
try:
|
| 309 |
+
with zipfile.ZipFile(zip_path, 'r') as zf:
|
| 310 |
+
names = zf.namelist()
|
| 311 |
+
|
| 312 |
+
# 查找 slices 目录
|
| 313 |
+
has_slices = any('slices/' in n for n in names)
|
| 314 |
+
has_textgrid = any('textgrid/' in n for n in names)
|
| 315 |
+
has_wav = any(n.endswith('.wav') for n in names)
|
| 316 |
+
has_lab = any(n.endswith('.lab') for n in names)
|
| 317 |
+
|
| 318 |
+
if not has_wav:
|
| 319 |
+
return False, "压缩包中未找到 .wav 音频文件", None
|
| 320 |
+
|
| 321 |
+
# 尝试从 meta.json 获取音源名称
|
| 322 |
+
source_name = None
|
| 323 |
+
if 'meta.json' in names:
|
| 324 |
+
try:
|
| 325 |
+
with zf.open('meta.json') as mf:
|
| 326 |
+
meta = json.load(mf)
|
| 327 |
+
source_name = meta.get('source_name')
|
| 328 |
+
except:
|
| 329 |
+
pass
|
| 330 |
+
|
| 331 |
+
# 如果没有 meta.json,从目录结构推断
|
| 332 |
+
if not source_name:
|
| 333 |
+
# 从 zip 文件名推断
|
| 334 |
+
source_name = Path(zip_path).stem.replace('_音源数据', '')
|
| 335 |
+
|
| 336 |
+
info_parts = []
|
| 337 |
+
if has_slices:
|
| 338 |
+
wav_count = len([n for n in names if 'slices/' in n and n.endswith('.wav')])
|
| 339 |
+
info_parts.append(f"切片: {wav_count} 个")
|
| 340 |
+
if has_textgrid:
|
| 341 |
+
tg_count = len([n for n in names if 'textgrid/' in n and n.endswith('.TextGrid')])
|
| 342 |
+
info_parts.append(f"TextGrid: {tg_count} 个")
|
| 343 |
+
|
| 344 |
+
info = " | ".join(info_parts) if info_parts else "有效的音源包"
|
| 345 |
+
|
| 346 |
+
return True, f"✅ {info}", source_name
|
| 347 |
+
|
| 348 |
+
except zipfile.BadZipFile:
|
| 349 |
+
return False, "无效的 zip 文件", None
|
| 350 |
+
except Exception as e:
|
| 351 |
+
return False, f"验证失败: {e}", None
|
| 352 |
+
|
| 353 |
+
|
| 354 |
+
def process_export_voicebank(
|
| 355 |
+
zip_file,
|
| 356 |
+
plugin_name: str,
|
| 357 |
+
max_samples: int,
|
| 358 |
+
naming_rule: str,
|
| 359 |
+
first_naming_rule: str,
|
| 360 |
+
progress=gr.Progress()
|
| 361 |
+
) -> Tuple[str, str, Optional[str]]:
|
| 362 |
+
"""
|
| 363 |
+
导出音源:上传音源包 → 解压 → 导出 → 打包下载
|
| 364 |
+
|
| 365 |
+
返回: (状态, 日志, 下载文件路径)
|
| 366 |
+
"""
|
| 367 |
+
logs = []
|
| 368 |
+
def log(msg):
|
| 369 |
+
logs.append(msg)
|
| 370 |
+
logger.info(msg)
|
| 371 |
+
|
| 372 |
+
# 验证输入
|
| 373 |
+
valid, msg, source_name = validate_voicebank_zip(zip_file)
|
| 374 |
+
if not valid:
|
| 375 |
+
return f"❌ {msg}", "", None
|
| 376 |
+
|
| 377 |
+
log(f"📦 {msg}")
|
| 378 |
+
log(f"📝 音源名称: {source_name}")
|
| 379 |
+
|
| 380 |
+
# 创建临时工作空间
|
| 381 |
+
workspace = create_temp_workspace()
|
| 382 |
+
log(f"🔧 创建工作空间")
|
| 383 |
+
|
| 384 |
+
try:
|
| 385 |
+
zip_path = zip_file.name if hasattr(zip_file, 'name') else str(zip_file)
|
| 386 |
+
|
| 387 |
+
# 解压音源包
|
| 388 |
+
progress(0.1, desc="解压音源包...")
|
| 389 |
+
bank_dir = os.path.join(workspace, "bank")
|
| 390 |
+
source_dir = os.path.join(bank_dir, source_name)
|
| 391 |
+
os.makedirs(source_dir, exist_ok=True)
|
| 392 |
+
|
| 393 |
+
success, msg = extract_zip(zip_path, source_dir)
|
| 394 |
+
if not success:
|
| 395 |
+
return f"❌ {msg}", "\n".join(logs), None
|
| 396 |
+
log(f"📂 已解压到工作目录")
|
| 397 |
+
|
| 398 |
+
# 检查目录结构,处理可能的嵌套
|
| 399 |
+
slices_dir = os.path.join(source_dir, "slices")
|
| 400 |
+
if not os.path.exists(slices_dir):
|
| 401 |
+
# 可能解压后有额外的一层目录
|
| 402 |
+
subdirs = [d for d in os.listdir(source_dir) if os.path.isdir(os.path.join(source_dir, d))]
|
| 403 |
+
if len(subdirs) == 1:
|
| 404 |
+
nested_dir = os.path.join(source_dir, subdirs[0])
|
| 405 |
+
if os.path.exists(os.path.join(nested_dir, "slices")):
|
| 406 |
+
# 移动内容到上层
|
| 407 |
+
for item in os.listdir(nested_dir):
|
| 408 |
+
shutil.move(os.path.join(nested_dir, item), source_dir)
|
| 409 |
+
os.rmdir(nested_dir)
|
| 410 |
+
|
| 411 |
+
# 执行导出
|
| 412 |
+
progress(0.3, desc="执行导出...")
|
| 413 |
+
log("\n" + "=" * 50)
|
| 414 |
+
log(f"【{plugin_name}】")
|
| 415 |
+
|
| 416 |
+
from src.export_plugins import load_plugins
|
| 417 |
+
plugins = load_plugins()
|
| 418 |
+
|
| 419 |
+
if plugin_name not in plugins:
|
| 420 |
+
return f"❌ 未找到插件: {plugin_name}", "\n".join(logs), None
|
| 421 |
+
|
| 422 |
+
plugin = plugins[plugin_name]
|
| 423 |
+
plugin.set_progress_callback(log)
|
| 424 |
+
|
| 425 |
+
options = {
|
| 426 |
+
"max_samples": max_samples,
|
| 427 |
+
"naming_rule": naming_rule,
|
| 428 |
+
"first_naming_rule": first_naming_rule,
|
| 429 |
+
"clean_temp": True
|
| 430 |
+
}
|
| 431 |
+
|
| 432 |
+
success, msg = plugin.export(source_name, bank_dir, options)
|
| 433 |
+
|
| 434 |
+
if not success:
|
| 435 |
+
return f"❌ 导出失败: {msg}", "\n".join(logs), None
|
| 436 |
+
|
| 437 |
+
log(f"✅ {msg}")
|
| 438 |
+
|
| 439 |
+
# 打包导出结果
|
| 440 |
+
progress(0.9, desc="打包结果...")
|
| 441 |
+
log("\n" + "=" * 50)
|
| 442 |
+
log("【打包结果】")
|
| 443 |
+
|
| 444 |
+
export_dir = os.path.join(workspace, "export", source_name, "simple_export")
|
| 445 |
+
|
| 446 |
+
# 如果导出目录不存在,尝试其他位置
|
| 447 |
+
if not os.path.exists(export_dir):
|
| 448 |
+
alt_export = os.path.join(os.path.dirname(bank_dir), "export", source_name, "simple_export")
|
| 449 |
+
if os.path.exists(alt_export):
|
| 450 |
+
export_dir = alt_export
|
| 451 |
+
|
| 452 |
+
if not os.path.exists(export_dir):
|
| 453 |
+
return "❌ 未找到导出结果", "\n".join(logs), None
|
| 454 |
+
|
| 455 |
+
zip_name = f"{source_name}_导出结果"
|
| 456 |
+
result_zip = create_zip(export_dir, zip_name)
|
| 457 |
+
|
| 458 |
+
if result_zip:
|
| 459 |
+
# 统计导出文件数
|
| 460 |
+
file_count = len([f for f in os.listdir(export_dir) if f.endswith('.wav')])
|
| 461 |
+
log(f"📦 已打包: {file_count} 个音频文件")
|
| 462 |
+
progress(1.0, desc="完成")
|
| 463 |
+
return "✅ 导出完成", "\n".join(logs), result_zip
|
| 464 |
+
else:
|
| 465 |
+
return "❌ 打包失败", "\n".join(logs), None
|
| 466 |
+
|
| 467 |
+
except Exception as e:
|
| 468 |
+
logger.error(f"导出失败: {e}", exc_info=True)
|
| 469 |
+
return f"❌ 处理失败: {e}", "\n".join(logs), None
|
| 470 |
+
|
| 471 |
+
finally:
|
| 472 |
+
cleanup_workspace(workspace)
|
| 473 |
+
|
| 474 |
+
|
| 475 |
+
# ==================== 构建界面 ====================
|
| 476 |
+
|
| 477 |
+
def create_cloud_ui():
|
| 478 |
+
"""创建云端 Gradio 界面"""
|
| 479 |
+
|
| 480 |
+
# 检查 MFA 状态
|
| 481 |
+
mfa_available = check_mfa_available()
|
| 482 |
+
mfa_status = "✅ MFA 已就绪" if mfa_available else "⚠️ MFA 不可用(将跳过对齐步骤)"
|
| 483 |
+
|
| 484 |
+
# 加载导出插件
|
| 485 |
+
from src.export_plugins import load_plugins
|
| 486 |
+
plugins = load_plugins()
|
| 487 |
+
plugin_names = list(plugins.keys()) if plugins else ["简单单字导出"]
|
| 488 |
+
|
| 489 |
+
with gr.Blocks(
|
| 490 |
+
title="人力V助手 (JinrikiHelper)",
|
| 491 |
+
theme=gr.themes.Soft()
|
| 492 |
+
) as app:
|
| 493 |
+
|
| 494 |
+
gr.Markdown("# 🎤 人力V助手 (JinrikiHelper)")
|
| 495 |
+
gr.Markdown("语音数据集处理工具 - 自动化制作语音音源库")
|
| 496 |
+
gr.Markdown("> ☁️ 云端版:上传音频 → 自动处理 → 下载结果")
|
| 497 |
+
|
| 498 |
+
with gr.Tabs():
|
| 499 |
+
# ==================== 制作音源页 ====================
|
| 500 |
+
with gr.Tab("🎵 制作音源"):
|
| 501 |
+
gr.Markdown("### 上传音频文件")
|
| 502 |
+
gr.Markdown("支持格式: WAV, MP3, FLAC, OGG, M4A")
|
| 503 |
+
|
| 504 |
+
audio_upload = gr.File(
|
| 505 |
+
label="上传音频文件",
|
| 506 |
+
file_count="multiple",
|
| 507 |
+
file_types=["audio"]
|
| 508 |
+
)
|
| 509 |
+
|
| 510 |
+
with gr.Row():
|
| 511 |
+
make_source_name = gr.Textbox(
|
| 512 |
+
label="音源名称",
|
| 513 |
+
placeholder="my_voice",
|
| 514 |
+
info="用于标识输出的音源包"
|
| 515 |
+
)
|
| 516 |
+
make_language = gr.Dropdown(
|
| 517 |
+
choices=CloudConfig.LANGUAGES,
|
| 518 |
+
value="chinese",
|
| 519 |
+
label="语言"
|
| 520 |
+
)
|
| 521 |
+
|
| 522 |
+
with gr.Row():
|
| 523 |
+
make_whisper = gr.Dropdown(
|
| 524 |
+
choices=list(CloudConfig.WHISPER_MODELS.keys()),
|
| 525 |
+
value="whisper-small",
|
| 526 |
+
label="Whisper 模型",
|
| 527 |
+
info="small 更快,medium 更准"
|
| 528 |
+
)
|
| 529 |
+
make_mfa_status = gr.Textbox(
|
| 530 |
+
label="MFA 状态",
|
| 531 |
+
value=mfa_status,
|
| 532 |
+
interactive=False
|
| 533 |
+
)
|
| 534 |
+
|
| 535 |
+
make_btn = gr.Button("🚀 开始制作", variant="primary", size="lg")
|
| 536 |
+
|
| 537 |
+
make_status = gr.Textbox(label="状态", interactive=False)
|
| 538 |
+
make_log = gr.Textbox(label="处理日志", lines=12, interactive=False)
|
| 539 |
+
|
| 540 |
+
gr.Markdown("### 下载结果")
|
| 541 |
+
make_download = gr.File(label="音源包下载", interactive=False)
|
| 542 |
+
|
| 543 |
+
gr.Markdown("""
|
| 544 |
+
> 💡 处理流程:
|
| 545 |
+
> 1. VAD 语音活动检测,自动切分音频
|
| 546 |
+
> 2. Whisper 语音识别,生成文本标注
|
| 547 |
+
> 3. MFA 强制对齐,生成音素级时间标注
|
| 548 |
+
> 4. 打包为 zip 供下载
|
| 549 |
+
""")
|
| 550 |
+
|
| 551 |
+
make_btn.click(
|
| 552 |
+
fn=process_make_voicebank,
|
| 553 |
+
inputs=[audio_upload, make_source_name, make_language, make_whisper],
|
| 554 |
+
outputs=[make_status, make_log, make_download]
|
| 555 |
+
)
|
| 556 |
+
|
| 557 |
+
# ==================== 导出音源页 ====================
|
| 558 |
+
with gr.Tab("📤 导出音源"):
|
| 559 |
+
gr.Markdown("### 上传音源包")
|
| 560 |
+
gr.Markdown("上传之前制作的音源压缩包(包含 slices 和 textgrid 目录)")
|
| 561 |
+
|
| 562 |
+
export_upload = gr.File(
|
| 563 |
+
label="上传音源包 (.zip)",
|
| 564 |
+
file_types=[".zip"]
|
| 565 |
+
)
|
| 566 |
+
|
| 567 |
+
export_info = gr.Textbox(
|
| 568 |
+
label="音���信息",
|
| 569 |
+
interactive=False,
|
| 570 |
+
placeholder="上传后显示音源信息"
|
| 571 |
+
)
|
| 572 |
+
|
| 573 |
+
# 上传后自动验证
|
| 574 |
+
def on_upload(file):
|
| 575 |
+
if file:
|
| 576 |
+
valid, msg, name = validate_voicebank_zip(file)
|
| 577 |
+
return msg
|
| 578 |
+
return ""
|
| 579 |
+
|
| 580 |
+
export_upload.change(
|
| 581 |
+
fn=on_upload,
|
| 582 |
+
inputs=[export_upload],
|
| 583 |
+
outputs=[export_info]
|
| 584 |
+
)
|
| 585 |
+
|
| 586 |
+
gr.Markdown("---")
|
| 587 |
+
gr.Markdown("### 导出设置")
|
| 588 |
+
|
| 589 |
+
export_plugin = gr.Dropdown(
|
| 590 |
+
choices=plugin_names,
|
| 591 |
+
value=plugin_names[0] if plugin_names else None,
|
| 592 |
+
label="导出插件"
|
| 593 |
+
)
|
| 594 |
+
|
| 595 |
+
with gr.Row():
|
| 596 |
+
export_max_samples = gr.Number(
|
| 597 |
+
label="每个拼音最大样本数",
|
| 598 |
+
value=10,
|
| 599 |
+
minimum=1,
|
| 600 |
+
maximum=1000
|
| 601 |
+
)
|
| 602 |
+
|
| 603 |
+
with gr.Row():
|
| 604 |
+
export_naming = gr.Textbox(
|
| 605 |
+
label="命名规则",
|
| 606 |
+
value="%p%%n%",
|
| 607 |
+
info="%p%=拼音, %n%=序号"
|
| 608 |
+
)
|
| 609 |
+
export_first_naming = gr.Textbox(
|
| 610 |
+
label="首个样本命名",
|
| 611 |
+
value="%p%",
|
| 612 |
+
info="第0个样本的特殊规则"
|
| 613 |
+
)
|
| 614 |
+
|
| 615 |
+
export_btn = gr.Button("📤 开始导出", variant="primary", size="lg")
|
| 616 |
+
|
| 617 |
+
export_status = gr.Textbox(label="状态", interactive=False)
|
| 618 |
+
export_log = gr.Textbox(label="处理日志", lines=10, interactive=False)
|
| 619 |
+
|
| 620 |
+
gr.Markdown("### 下载结果")
|
| 621 |
+
export_download = gr.File(label="导出结果下载", interactive=False)
|
| 622 |
+
|
| 623 |
+
gr.Markdown("""
|
| 624 |
+
> 💡 导出说明:
|
| 625 |
+
> - 从 TextGrid 提取每个汉字/音节的时间边界
|
| 626 |
+
> - 按拼音/罗马音分类,选取最佳样本
|
| 627 |
+
> - 导出为适配其他软件的音源格式
|
| 628 |
+
""")
|
| 629 |
+
|
| 630 |
+
export_btn.click(
|
| 631 |
+
fn=process_export_voicebank,
|
| 632 |
+
inputs=[
|
| 633 |
+
export_upload, export_plugin,
|
| 634 |
+
export_max_samples, export_naming, export_first_naming
|
| 635 |
+
],
|
| 636 |
+
outputs=[export_status, export_log, export_download]
|
| 637 |
+
)
|
| 638 |
+
|
| 639 |
+
# ==================== 关于页 ====================
|
| 640 |
+
with gr.Tab("ℹ️ 关于"):
|
| 641 |
+
gr.Markdown("""
|
| 642 |
+
## 人力V助手 (JinrikiHelper)
|
| 643 |
+
|
| 644 |
+
语音数据集处理工具,用于自动化制作语音音源库。
|
| 645 |
+
|
| 646 |
+
### 功能特点
|
| 647 |
+
|
| 648 |
+
- **VAD 切片**: 使用 Silero VAD 自动检测语音片段
|
| 649 |
+
- **语音识别**: 使用 Whisper 模型转录文本
|
| 650 |
+
- **强制对齐**: 使用 MFA 生成音素级时间标注
|
| 651 |
+
- **智能导出**: 按拼音分类,选取最佳样本
|
| 652 |
+
|
| 653 |
+
### 支持语言
|
| 654 |
+
|
| 655 |
+
- 中文(普通话)
|
| 656 |
+
- 日语
|
| 657 |
+
|
| 658 |
+
### 使用流程
|
| 659 |
+
|
| 660 |
+
1. **制作音源**: 上传原始音频 → 自动处理 → 下载音源包
|
| 661 |
+
2. **导出音源**: 上传音源包 → 选择导出格式 → 下载导出结果
|
| 662 |
+
|
| 663 |
+
---
|
| 664 |
+
|
| 665 |
+
**作者**: TNOT | **协议**: MIT
|
| 666 |
+
|
| 667 |
+
本工具集成 Montreal Forced Aligner (MIT License)
|
| 668 |
+
""")
|
| 669 |
+
|
| 670 |
+
return app
|
| 671 |
+
|
| 672 |
+
|
| 673 |
+
def main():
|
| 674 |
+
"""云端入口"""
|
| 675 |
+
app = create_cloud_ui()
|
| 676 |
+
app.launch(
|
| 677 |
+
server_name="0.0.0.0",
|
| 678 |
+
server_port=7860,
|
| 679 |
+
share=False,
|
| 680 |
+
show_error=True
|
| 681 |
+
)
|
| 682 |
+
|
| 683 |
+
|
| 684 |
+
if __name__ == "__main__":
|
| 685 |
+
main()
|