REAL Video Enhancer: AI-Powered Video Interpolation, Upscaling, and Denoising

Introduction

REAL Video Enhancer is a comprehensive, open-source application that redefines video enhancement. It serves as a redesigned and improved version of the original Rife ESRGAN App for Linux, now offering robust functionalities across Windows, Linux, and macOS. This project aims to provide a superior alternative to outdated software like Flowframes or enhancr, making advanced video processing accessible.

Key features include:

Cross-platform support for Windows, Ubuntu 22.04+, and MacOS 15+ (arm/x86).
Discord RPC integration for system and Flatpak Discord packages.
Intelligent scene change detection to maintain sharp transitions.
A live preview displaying the latest rendered frame.
Optimized inference across various GPUs using TensorRT and NCNN.

Installation

Please note: This project is currently undergoing a significant refactor, and the current version may not receive updates for a while. However, the existing stable version remains functional.

To get started with REAL Video Enhancer, you can clone the repository and build it.

Cloning the Repository:

For the nightly version:

git clone --recurse-submodules https://github.com/TNTwise/REAL-Video-Enhancer

For the stable 2.4.1 version:

git clone --recurse-submodules https://github.com/TNTwise/REAL-Video-Enhancer --branch 2.4.1

Building the Application:

REAL Video Enhancer supports multiple build methods and Python versions.
Supported Python versions: 3.10, 3.11, 3.12.
Supported build methods: pyinstaller (recommended for Windows/Mac), cx_freeze (recommended for Linux), nuitka (experimental).

To build, navigate to the cloned directory and run:

python3 build.py --build BUILD_OPTION --copy_backend

Replace BUILD_OPTION with your preferred method (e.g., pyinstaller or cx_freeze).

Examples

REAL Video Enhancer excels at transforming video content through various AI-powered processes. You can use it to:

Interpolate Frames: Smooth out videos by generating intermediate frames, effectively converting 24FPS video to 48FPS or higher for a more fluid viewing experience.
Upscale Resolution: Enhance video resolution, for instance, upscaling a 1920x1080 video to 3840x2160 (4K) using advanced models like Real-ESRGAN and SPAN.
Decompress and Denoise: Improve video clarity by reducing compression artifacts and removing unwanted noise, making older or lower-quality footage look significantly better.

The application supports a wide array of interpolation models (e.g., RIFE, GMFSS, IFRNet) and upscale models (e.g., 4x-SPANkendata, 2x-AnimeJaNai V2). For a quick demonstration, refer to the project's demo screenshot showcasing its capabilities.

If you prefer to experiment without a local installation, a Colab Notebook is available for cloud-based processing.

Why Use It

Choosing REAL Video Enhancer means opting for a cutting-edge, versatile, and user-friendly solution for video enhancement.

Cross-Platform Compatibility: Enjoy seamless operation on Linux, Windows, and macOS, catering to a broad user base.
Advanced AI Integration: Leverage state-of-the-art AI models for superior frame interpolation, upscaling, decompression, and denoising results.
Optimized Performance: Benefit from efficient inference backends like TensorRT, PyTorch, and NCNN, ensuring faster processing on compatible GPUs.
Modern Alternative: Move beyond outdated video processing tools with an actively developed and feature-rich application.
Open Source: As an open-source project, it fosters community contributions and transparency, allowing users to inspect, modify, and improve the software.
User-Friendly Interface: The application provides a graphical user interface (GUI) that simplifies complex video enhancement tasks, making them accessible to users of all skill levels.

REAL Video Enhancer: AI-Powered Video Interpolation, Upscaling, and Denoising

Summary

Repository Information

Topics

Use at your own risk