A powerful audio processing tool that converts audio and video files to text.

• Support for multiple audio and video formats
• Automatic audio extraction from video
• Smart segmentation for long audio files
• Multiple ASR service polling with auto-retry
• Real-time progress display
• File monitoring mode
• Flexible configuration options

1. Clone the repository:

git clone [repository-url] cd segement_audio

2. Install dependencies:

pip install -r requirements.txt

3. Install FFmpeg (for video processing):

• Windows: Download from https://ffmpeg.org/download.html and add to system PATH
• Linux: sudo apt-get install ffmpeg
• MacOS: brew install ffmpeg

Basic usage:

python main.py -m /path/to/media -o /path/to/output

Arguments:

• -c, --config: Configuration file path (default: config.json)
• -m, --media-folder: Media folder path
• -o, --output-folder: Output folder path
• -w, --watch: Enable watch mode
• --no-video: Don't process video files
• --extract-only: Only extract audio
• --no-progress: Don't show progress bars
• --no-timestamps: Don't include timestamps
• --debug: Enable debug mode

Configure options in config.json:

{ "media_folder": "./media", # Media folder path "output_folder": "./output", # Output folder path "max_retries": 3, # Maximum retry attempts "max_workers": 4, # Maximum concurrent worker threads "use_jianying_first": true, # Prioritize Jianying ASR "use_kuaishou": true, # Use Kuaishou ASR "use_bcut": true, # Use Bcut ASR "format_text": true, # Format output text "include_timestamps": true, # Include timestamps "show_progress": true, # Show progress bars "process_video": true, # Process video files "extract_audio_only": false, # Only extract audio "watch_mode": false, # Watch mode "segment_length": 30, # Audio segment length (seconds) "max_segment_length": 2000, # Maximum text segment length "min_segment_length": 10, # Minimum text segment length "retry_delay": 1.0, # Retry delay (seconds) "log_level": "INFO", # Logging level "log_file": "./output/audio_processing.log" # Log file path }

1. Basic usage:

python main.py -m ./media -o ./output

2. Watch mode:

python main.py -m ./media -o ./output -w

3. With custom config:

python main.py -c my_config.json

4. Extract audio only:

python main.py -m ./media -o ./output --extract-only

Text output example:

# example.mp3 # Processed: 2023-12-25 14:30:00 # Total segments: 10 # Successfully transcribed: 10 [00:00 - 00:30] First transcribed text segment... [00:30 - 01:00] Second transcribed text segment... 

• Automatic retry of failed recognition tasks
• Detailed error logging
• Process can be interrupted and progress saved
• Supports resuming from interruption

1. FFmpeg is required for video processing
2. Python 3.7+ recommended
3. Ensure sufficient disk space for temporary files
4. Watch mode runs until manually stopped

MIT License

要冻结项目依赖，请运行：

# 将依赖冻结到默认的requirements.txt文件 python freeze_deps.py freeze # 或指定输出文件 python freeze_deps.py freeze --file requirements_dev.txt

要从冻结的依赖文件中安装包：

# 从默认的requirements.txt文件安装 python freeze_deps.py install # 或从指定文件安装 python freeze_deps.py install --file requirements_dev.txt

你也可以直接使用pip命令：

# 冻结依赖 pip freeze > requirements.txt # 安装依赖 pip install -r requirements.txt

如果你发现某些包(如watchdog)没有出现在freeze生成的requirements.txt中，可能有以下原因:

1. 环境不匹配: 确保你在正确的Python环境中运行freeze命令
2. 包未正确安装: 某些包可能不是通过pip安装的
3. 虚拟环境问题: 确保你在激活了正确的虚拟环境后再运行命令

我们提供了一个增强版的依赖冻结脚本，可以确保关键包被包含:

# 运行检查并冻结依赖 python check_and_freeze_deps.py

这个脚本会:

• 检查watchdog是否已安装
• 提供安装选项(如果未安装)
• 确保watchdog被添加到requirements.txt

如果您仍然遇到问题，可以手动添加watchdog到您的requirements.txt:

watchdog==x.y.z # 替换x.y.z为您需要的版本