NVIDIA VFX Python Samples

This repository contains sample applications for the nvidia-vfx Python bindings of the NVIDIA Video Effects (VFX) SDK. The Video Super Resolution sample is a Python reference application that demonstrates real-time video upscaling: it reads an input video file, upscales each frame using GPU-accelerated AI models to improve sharpness and reduce compression artifacts, and writes the enhanced result to a new output file based on command-line options.

Requirements

• Python: 3.10 or later
• GPU: NVIDIA GPU with Tensor Cores (Turing, Ampere, Ada, Blackwell, or Hopper architecture)
• GPU driver:
  • Windows: 570.65 or later (for Tesla Compute Cluster (TCC) devices, 595 or later is required)
  • Linux: 570.190+, 580.82+, or 590.44+
• Git: https://git-scm.com/downloads
• Git LFS: https://git-lfs.com/ (required for sample video assets)
• OS:
  • Windows: 64-bit Windows 10 or later
  • Linux: Ubuntu 20.04, 22.04, 24.04, Debian 12, or RHEL 8/9

Setup

Clone the repository

git clone git@github.com:NVIDIA-Maxine/nvidia-vfx-python-samples.git        # Using SSH, or
# git clone https://github.com/NVIDIA-Maxine/nvidia-vfx-python-samples.git  # Using HTTP
cd nvidia-vfx-python-samples

Initialize Git LFS so sample videos are downloaded correctly:

git lfs install
git lfs pull

Create a virtual environment and install dependencies

python -m venv .venv
source .venv/bin/activate        # Linux
# .venv\Scripts\Activate.ps1     # Windows PowerShell
pip install -r requirements.txt

Running the application

From the repository root (with your virtual environment activated):

# Default: 2× upscale, HIGH quality, using bundled sample video
python video_super_resolution.py

# Custom input and output
python video_super_resolution.py -i path/to/input.mp4 -o path/to/output.mp4

# 4× upscale with ULTRA quality
python video_super_resolution.py --scale 4 --quality ULTRA

# Same-resolution denoise (no upscaling)
python video_super_resolution.py --scale 1 --quality DENOISE_HIGH

Video Super Resolution Sample — Command-Line Reference

Argument	Description
-i, --input	The input video file path. Default: assets/Drift_RUN_Master_Custom.mp4 (relative to the script location)
-o, --output	The output video file path. Default: output/sample_sr.mp4 (relative to the script location)
--scale	The scale factor. Choices: 1, 2, 3, 4. Default: 2. Values 2, 3, and 4 upscale both width and height (for example, 1920x1080 → 3840x2160 at 2×), with higher factors demanding more GPU memory and compute. A value of 1 keeps the original resolution and enables same‑resolution cleanup modes (denoise/deblur) instead of upscaling.
--quality	The processing mode / quality level. For --scale ≥ 2, choices: BICUBIC, LOW, MEDIUM, HIGH, ULTRA (default: HIGH), where BICUBIC is a fast non‑AI baseline, LOW/MEDIUM favor speed, and HIGH/ULTRA maximize detail enhancement and artifact reduction. When --scale 1 is used, additional choices: DENOISE_LOW, DENOISE_MEDIUM, DENOISE_HIGH, DENOISE_ULTRA (noise/compression artifact removal) and DEBLUR_LOW, DEBLUR_MEDIUM, DEBLUR_HIGH, DEBLUR_ULTRA (sharpening for soft or blurry footage).
-h, --help	Display help information for the command.

Links

• nvidia-vfx Python package on PyPI
• nvidia-vfx Python bindings guide

Note: This project is currently not accepting contributions.