🎨 Anime Enhancement Suite

Cross-platform Python toolchain for high-quality anime video restoration using neural networks.
Upscales, denoises, and interpolates anime videos — from old 360p/480p to sharp 4K+ — while preserving original audio and optimizing for performance via batching, parallelism, and lightweight binaries.

---

🔍 What It Does

• 🎨 Upscaling: Enhances frames with Real-ESRGAN
• 🧹 Denoising: Removes noise/artifacts with waifu2x-ncnn-vulkan
• 🎞️ Interpolation: Smooths motion & increases FPS with RIFE-ncnn-vulkan
• 🔊 Audio preservation: Extracts, encodes, and syncs back audio tracks
• ⚡ Optimized performance: Frame batching + threading/multiprocessing
• 🖥️ Cross-platform binaries included: Works out-of-the-box on Windows, Linux, macOS

---

🚀 Quick Start

✅ Requirements

• Python 3.12+
• ffmpeg installed and available in system PATH
• Basic dependencies managed by poetry

🧪 Installation

git clone https://github.com/Olexasha/anime_enhancement.git
cd anime_enhancement
poetry install

▶️ Usage

1. Place your video file in data/input_video;
2. Set video filename in src/config/settings.py under ORIGINAL_VIDEO
3. (Optional) Edit .envor settings (batch size, scale factor, FPS, denoise strength)
4. Run enhancement:

python main.py
# or if inside poetry environment:
poetry run python main.py

---

📦 Example Workflow

Input: data/input_video/naruto_war.mp4

Pipeline:

1. Extract audio
2. Extract frames
3. Apply denoise → upscale → interpolation
4. Merge video parts
5. Reattach synced audio

Output:

• data/output_video/naruto_war_enhanced.mp4

---

🧩 Features

• 🔁 Batch-based frame processing (low memory footprint)
• 🧵 IO-bound multithreading + CPU/GPU parallelism
• ⚙️ Custom Real-ESRGAN execution per OS
• 🎛️ Configurable stages: enable/disable upscale, denoise, interpolation
• ⚙️ Cross-OS binaries for ESRGAN, waifu2x, RIFE
• 💾 Automatic cleanup of temp files
• 📤 GUI and CLI (in progress)

---

⚙️ Configuration

Main settings:

• src/config/settings.py — core parameters (video name, batch size, scale factor, etc.)
• .env — for optional overrides and system-level customization

---

🧠 Tech Stack

• Python 3.12+
• ffmpeg — audio/video I/O
• Real-ESRGAN — upscaling
• Waifu2x — denoising
• RIFE — frame interpolation
• threading, multiprocessing, asyncio — parallel execution
• poetry — dependency management

---

🖼 Directory Structure

anime_enhancement/
├── data/
│   ├── input_video/              # Input video files
│   ├── output_video/             # Final enhanced videos
│   ├── audio/                    # Extracted audio files
│   ├── tmp_video/                # Temporary merging outputs
│   ├── default_frame_batches/    # Extracted frames
│   ├── upscaled_frame_batches/   # Enhanced frames
│   └── video_batches/            # Reconstructed video parts
├── src/
│   ├── audio/                    # Audio extraction/merging logic
│   ├── config/                   # Settings and constants
│   ├── files/                    # File management utilities
│   ├── frames/                   # Frame extraction and enhancement
│   ├── video/                    # Video processing logic
│   ├── utils/                    # Logging and Real-ESRGAN binaries
│   ├── interfaces/               # GUI and API (under development)
│   └── tests/                    # Pytest config (tests TBD)
├── pyproject.toml                # Poetry dependencies
├── main.py                       # Entry point
└── .env                          # Optional environment variables

---

📋 Future Roadmap

• Add support for alternative upscaling models
• Add denoising neural network (e.g. waifu2x)
• Add motion interpolation (e.g. RIFE)
• Argparse CLI mode
• Complete GUI interface
• REST API for remote processing
• Upload to HuggingFace or PyPI

---

🧪 Testing

Planned to use pytest. Config file exists under src/tests/pytest.ini
Run tests (once added) with:

poetry run pytest

---

💬 Contributing

Contributions welcome!

• Fork → PR
• File issues & feature requests
• Test across OSes & report bugs

---

📜 License

MIT License — see LICENSE file for full text.

---

🧠 Credits

• Waifu2x by nihui
• Real-ESRGAN by Xintao
• RIFE by nihui
• Inspired by anime remastering communities