TNTwise /REAL-Video-Enhancer

About REAL-Video-Enhancer

REAL Video Enhancer is a desktop app for boosting frame rate and resolution. It puts frame interpolation, upscaling, denoising and decompression models behind one window. Expect to need a decent GPU: TensorRT wants an NVIDIA RTX card with enough VRAM. Two warnings: the Windows build can be falsely flagged as a trojan, and ROCm support is buggy.

Best for: Editors who need a local app to smooth, upscale or clean up video files without writing a script pipeline.

• Raises frame rate, for example 24 to 48 FPS, with RIFE, GMFSS and IFRNet
• Upscales video, for example 1920 to 3840, with SPAN, OpenProteus and AnimeSR
• Cleans up files with DeH264 decompression and DRUnet or DnCNN denoising
• Detects scene changes so sharp cuts stay sharp
• Shows a live preview of the last rendered frame
• Runs on TensorRT, PyTorch (CUDA 12.6 or ROCm 6.2) or NCNN Vulkan

1. 1
   Clone the repo with its submodules: the command below

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

   Copy
2. 2
   For the stable release instead, add --branch 2.4.1 to that clone command
3. 3
   Check you have Python 3.10, 3.11 or 3.12 installed
4. 4
   Build it with the command below

   python3 build.py --build BUILD_OPTION --copy_backend

   Copy
5. 5
   Or skip the build: download the prebuilt release from Steam, or run the Colab notebook