Overview

NVIDIA Video Super Resolution (VSR) NIM is an AI inference microservice for upscaling video to higher resolutions. It supports two runtime modes with separate deployment artifacts:

• ST 2110 video workflows: SMPTE ST 2110-20 uncompressed video workflows on NVIDIA Holoscan for Media.

• Compressed video workflows: gRPC, RTP/UDP, and Cloud File workflows for encoded video, controlled through the VSR gRPC API.

The deployment pages keep site-specific values such as secret names, namespaces, and node names as placeholders. Use the following image repositories as the release defaults when an image repository override is required.

The ST 2110 video workflow charts use the H4M VSR NIM image:

nvcr.io/nim/nvidia/vsr-h4m-nim:1.0.10

The compressed-video workflow uses the standalone VSR NIM image:

nvcr.io/nim/nvidia/vsr-nim:1.0.10

Choose a Mode

Use the following table to choose where to start.

Mode	Use When	Start Here
ST 2110 video workflows	You want to deploy VSR into a Holoscan for Media pipeline for live SMPTE ST 2110-20 uncompressed video.	ST 2110 Video Workflows
Compressed video workflows	You want to run the NIM as a container and send encoded-video requests with gRPC, RTP/UDP, or Cloud File inputs and outputs.	Compressed Video Workflows

Shared Requirements

VSR NIM 1.0.10 requires NVIDIA Driver 590.33 or later, CUDA 13.1 or later, and an Ada-generation or later NVIDIA GPU. Additional requirements are mode-specific and are documented in the support matrix for each mode:

• ST 2110 Video Workflow Support Matrix

• Compressed Video Workflow Support Matrix