Installation#

Prerequisites (NGC authentication, secrets, high-speed network) and Holoscan for Media platform setup are covered in Getting Started. For strict NUMA clusters and the schedulerName setting, see Scheduler Requirements for NUMA-aware Clusters.

Pull Helm Charts#

Add the Helm repository and pull the charts. Use chart version 1.0.0 (listed in the NGC catalog for LipSync on Holoscan for Media).

export NGC_API_KEY=<PASTE_API_KEY_HERE>

helm repo add nim-repo \
  https://helm.ngc.nvidia.com/nim/nvidia/ \
  '--username=$oauthtoken' --password=$NGC_API_KEY

helm repo update

Pull the charts:

helm pull nim-repo/nvidia-lipsync-h4m-sample --version 1.0.0
helm pull nim-repo/nvidia-lipsync-h4m-service --version 1.0.0
helm pull nim-repo/nvidia-lipsync-h4m-operator --version 1.0.0

Deployment Options#

The following pages describe Helm-based installation for LipSync on Holoscan for Media:

  • NIM Service Chart (nvidia-lipsync-h4m-service): NIM service deployment only, for integration into a custom pipeline.

  • Kubernetes Operator Chart (nvidia-lipsync-h4m-operator): Operator controller and NvidiaLipsyncMediaFunction custom resource.

  • End-to-End Demo Chart (nvidia-lipsync-h4m-sample): Sender, NIM service, and receiver deployed as a single Helm release for evaluation.

Helm values and operator custom resource fields are summarized in Configuration Reference.