Skip to navigation

Maxine NIM Overview

View as MarkdownOpen in Claude

NVIDIA Maxine NIMs are a collection of GPU-accelerated inference microservices that bring AI-powered video and audio effects behind consistent service interfaces. Each NIM packages a model as a self-hosted, easily deployable container, so you can add effects such as eye contact, relighting, lip sync, speech enhancement and noise removal to live and recorded media workflows. Holoscan for Media variants deploy the same capabilities into live production environments over SMPTE ST 2110.

Use Cases

Maxine NIMs span video enhancement, video analysis, speech-driven animation, audio enhancement and live production.

Video Enhancement

Use these NIMs when you need to improve or restyle video:

Benefits: Self-hosted GPU-accelerated containers, gRPC service interfaces

Video Analysis

Use these NIMs when you need to understand what is in a video:

Benefits: Per-frame and per-stream analysis, self-hosted containers, gRPC service interfaces

Speech-Driven Animation

Use these NIMs when you need facial movement that follows speech:

Benefits: Speech audio combined with video or a portrait photo, self-hosted containers

Audio Enhancement

Use these NIMs when you need cleaner speech audio:

Benefits: Self-hosted GPU-accelerated containers, gRPC service interfaces

Holoscan for Media

Use these NIMs when you need Maxine effects in live production over SMPTE ST 2110:

Benefits: Helm chart and Kubernetes operator deployment, SMPTE ST 2110 and NMOS integration