NotY Caption Generator AI v7.1 发布：基于 OpenAI Whisper 的视频字幕生成工具

🎬 NotY Caption Generator AI v7.1

A powerful command-line AI tool that generates professional subtitles/captions for videos and audio files using OpenAI's Whisper, with advanced vocal separation using Spleeter, YouTube download support, and smart subtitle formatting.

📋 Table of Contents

• Features
• System Requirements
• Quick Installation
• What's New in v7.1
• How to Use
• Project Structure
• Build from Source
• Configuration
• Troubleshooting
• Performance
• Support
• License

✨ Features

Core Functionality

• 🎯 AI-Powered Transcription - Uses OpenAI Whisper for high accuracy with word-level timestamps
• 🎤 Professional Vocal Separation - Spleeter integration (2/4/5 stems) with local models support
• 📺 YouTube Support - Download and caption any YouTube video
• 🌍 Multi-Language - 7 languages + Auto Detect (English, Hindi, Japanese, Spanish, Korean, Chinese, Russian)
• 🔄 Translation Mode - Transcribe and translate to English simultaneously
• 📝 Smart Subtitle Formatting - Word/Character/Auto line breaks with intelligent sentence detection
• ⚡ Audio Chunking - Automatic 10-second chunk processing for better performance

Technical Improvements (v7.1)

• 🚀 60-70% Faster Startup - Nuitka compiled executable
• 📦 Modular Architecture - Clean separation of core, utils, and UI modules
• 🗑️ Automatic Cache Cleanup - No leftover files on exit
• 🔧 Enhanced Text Cleaning - Removes hallucinations and repeated words
• 💾 Reduced Memory Usage - 25% less RAM consumption

📋 System Requirements

🚀 Quick Installation

Option 1: Installer (Recommended)

Download NotYCaptionGenAI_Installer_v7.1.exe and run:

NotYCaptionGenAI_Installer_v7.1.exe

The installer will:

• Install to C:\NotYCaptionGenAI
• Copy all models (Whisper + Spleeter)
• Create Desktop and Start Menu shortcuts
• Add to Windows "Send To" menu
• Register in Add/Remove Programs

Option 2: Portable

Download NotYCaptionGenAI_Portable_v7.1.zip and extract anywhere.

🆕 What's New in v7.1

Major Changes

✨ COMPLETE MODULAR ARCHITECTURE REWRITE
   • Codebase restructured into clean module hierarchy
   • Separation of concerns: core, utils, UI modules
   • Better maintainability and scalability

✨ DUAL BUILD SYSTEM IMPLEMENTED
   • Nuitka for faster executable startup (60-70% faster)
   • PyInstaller for reliable installer creation
   • _packages_ folder for bundled dependencies

✨ AUDIO CHUNKING SYSTEM
   • Automatic audio splitting into 10-second chunks
   • 1-second overlap between chunks for seamless transcription
   • Intelligent merging of chunk results

Fixed Issues

🐛 RESOLVED
   ✓ TensorFlow initialization errors
   ✓ Spleeter 'multiprocess_count' parameter error
   ✓ Repeated word hallucination (e.g., "अपने अपने अपने")
   ✓ Cache folder not deleting on application close
   ✓ _vocals suffix removed from output filenames
   ✓ Language auto-detection in Normal mode
   ✓ Import error with Colors class

🎮 How to Use

Method 1: Run from Start Menu

Click Start → NotY Caption Generator AI

Method 2: Right-click any file

Right-click any video/audio file → Send To → NotYCaptionGenAi

Method 3: Command Line

NotYCaptionGenAI.exe "C:\path\to\video.mp4"

Interactive Menu Flow

1. Choose platform: YouTube or Local File
2. Select vocal separation (none/2stems/4stems/5stems)
3. Choose Whisper model (tiny/base/small/medium/large)
4. Select mode: Normal or Translate
5. Pick language (or Auto Detect)
6. Choose line break type (Auto/Words/Letters)
7. Generate captions!

📁 Project Structure

NotYCaptionGenAi/
├── noty_caption_gen.py          # Main entry point
├── requirements.txt              # Dependencies (locked versions)
├── README.md                     # This file
├── LICENSE                       # LGPL-3.0
│
├── Builder/                      # Build system
│   ├── build_nuitka.py          # Nuitka build script (FAST)
│   ├── build_installer.py       # PyInstaller installer builder
│   └── setup.py                 # Complete build orchestrator
│
├── src/                          # Modular source code
│   ├── core/                    # Core functionality
│   │   ├── transcriber.py       # Whisper transcription
│   │   ├── vocal_separator.py   # Spleeter/FFmpeg separation
│   │   ├── audio_processor.py   # Audio extraction & chunking
│   │   └── subtitle_generator.py # SRT generation
│   ├── utils/                   # Utilities
│   │   ├── colors.py            # ANSI color codes
│   │   ├── cleanup.py           # Temp file cleanup
│   │   ├── cache_manager.py     # Database caching
│   │   └── file_handlers.py     # File dialogs
│   └── ui/                      # User interface
│       └── menu.py              # Console menu system
│
├── _packages_/                   # Bundled dependencies (created by installer)
│
├── ffmpeg/                       # FFmpeg binaries
│   ├── ffmpeg.exe
│   ├── ffplay.exe
│   └── ffprobe.exe
│
├── models/                       # Whisper model files
│   ├── tiny.pt                   # 75 MB - Fastest
│   ├── base.pt                   # 150 MB - Balanced
│   └── small.pt                  # 500 MB - Good
│
├── pretrained_models/            # Spleeter models
│   └── 5stems/                   # 5-stem separation model
│       ├── checkpoint
│       ├── model.data-00000-of-00001
│       ├── model.index
│       └── model.meta
│
└── resources/                    # Icons & assets
    ├── app.ico
    └── logo.ico

📦 Included Models

Whisper Models (Speech Recognition)

Spleeter Models (Vocal Separation)

🛠️ Build from Source

Prerequisites

# Python 3.10 (required for TensorFlow/Spleeter)
python --version  # Should be 3.10.x

# Install build tools
pip install nuitka pyinstaller

# Install dependencies
pip install -r requirements.txt

Build Commands

# Complete build (executable + installer)
python Builder/setup.py

# Build only executable with Nuitka (faster)
python Builder/build_nuitka.py

# Build only installer (requires executable)
python Builder/build_installer.py

Build Output

⚙️ Configuration

Supported File Formats

Language Codes

Environment Variables

# Optional performance tuning
set OMP_NUM_THREADS=4
set TF_CPP_MIN_LOG_LEVEL=2

📊 Performance Metrics

Speed Comparison (v6.1 vs v7.1)

Model Performance (per minute of audio)

🐛 Troubleshooting

Issue: "TensorFlow not available"

Solution:

• Ensure Python 3.10 is installed
• Run: pip install tensorflow==2.10.0
• Check: python -c "import tensorflow as tf; print(tf.__version__)"

Issue: "Spleeter model not found"

Solution:

• Models are included in pretrained_models/ folder
• If missing, Spleeter will download on first use (~200 MB)
• Ensure write permissions to installation directory

Issue: "Out of memory"

Solution:

• Use smaller Whisper model (tiny or base)
• Disable vocal separation
• Close other applications
• Use 2stems instead of 5stems

Issue: "FFmpeg not found"

Solution:

• FFmpeg is bundled with the installer
• For portable version, ensure ffmpeg folder exists
• Or install FFmpeg system-wide

Issue: "Repeated words in output"

Solution:

• v7.1 includes automatic hallucination removal
• Use "Auto" line break mode
• Try a larger Whisper model for better accuracy

Issue: "Cache folder not deleting"

Solution:

• v7.1 automatically cleans cache on exit
• Manual cleanup: delete C:\NotYCaptionGenAI\cache

🤝 Contributing

Contributions are welcome! Please:

1. Fork the repository
2. Create a feature branch
3. Submit a pull request

Development Setup

git clone https://github.com/NotY215/NotYCaptionGenAi.git
cd NotYCaptionGenAi
python -m venv venv
venv\Scripts\activate
pip install -r requirements.txt

🙏 Acknowledgments

• OpenAI Whisper - Speech recognition technology
• Deezer Spleeter - Vocal separation engine
• yt-dlp - YouTube downloading
• FFmpeg - Audio processing backbone
• TensorFlow - Machine learning framework

📈 Version History

⭐ Star the Project

If you find this tool useful, please consider starring the repository on GitHub!

Made with ❤️ by NotY215

Last Updated: April 2026