NVIDIA Holoscan for Media
NVIDIA Holoscan for Media is a unified, scalable, IP-based platform that brings real-time AI into live production for broadcast, news, and sports. It lets teams run multiple production and AI applications on shared NVIDIA-accelerated infrastructure using flexible, software-defined workflows.
Developers across the ecosystem can leverage the platform to build and deploy AI applications that connect to uncompressed media feeds with minimal latency.
Key Benefits of Holoscan for Media for Developers
Streamline AI Application Development in Live Media
Connects AI Applications to Live Pipelines
Framework integrates DeepStream and Rivermax to connect AI applications and services as well as AI development software to uncompressed video streams
Supports Cloud-Native Development
Provides reference applications that guide developers on building and deploying their solutions in a cloud-native environment.
Solves the Complexities of Uncompressed Media
Ensures compliance with industry standards for uncompressed video like SMPTE 2110 enabled by NVIDIA Rivermax, as well as NMOS and PTP.
What’s New In Holoscan for Media?
Easy-to-use microservices and SDKs designed for secure, reliable, high-performance deployment across clouds, data centers, and edge:
Active Speaker Detection
ASD SMPTE ST 2110 brings multi‑speaker detection and identification to live broadcast workflows over IP video. It enables real‑time speaker tagging within NVIDIA Holoscan for Media.
Studio Voice NIM
Studio Voice SMPTE ST 2110 brings studio‑quality speech enhancement to live broadcast audio pipelines. It supports professional IP‑based media workflows using standard input equipment.
Lip Sync
The Lip‑Sync SMPTE ST 2110 NIM synchronizes lip movements with speech in live, IP‑based broadcast video pipelines. It is designed for real‑time dubbing workflows in NVIDIA Holoscan for Media environments.
RTX Video Super Resolution
RTX Video Super Resolution upscales 16:9 video from 480p to as high as 8K using AI, with user controls for sharpness, blur, denoising, and hallucination limits. The model can be fine‑tuned to source content and runs within NVIDIA AI for Media. Also available as a Python Wheel.
Platform Architecture
Streamline AI Application Development in Live Media
Subscribe to Receive Updates
Get the latest updates about Holoscan for Media and our other live media solutions straight to your inbox.