Uploaded May 2026 | Updated September 2026, 2 weeks ago
Organizations increasingly rely on video to capture critical information—yet extracting meaningful insights from massive amounts of footage remains a challenge. NVIDIA Metropolis Blueprint for video search and summarization (VSS) overcomes this hurdle by transforming hours of video into instantly searchable, actionable intelligence.
With advanced indexing, scalable architecture, and real-time analytics, VSS helps enterprises monitor operations, detect trends, and make informed decisions faster than ever. VSS redefines video search, analysis, and reporting through next-generation multimodal intelligence and modular design.
Now with NVIDIA NemoClaw, an open source stack that adds privacy and security controls to OpenClaw, you can deploy video agents using VSS skills.
What you'll learn:
Build a video agent with VSS skills on Brev launchable and deploy a search profile for semantic fusion search with NemoClaw.
Use VSS skills with NemoClaw to integrate agents into Slack to search videos and review clips from chat.
Configure agents to generate real-time VLM-powered alerts from a live streaming video.
Understand the end‑to‑end pipeline to build agents and monitoring workflows with the OpenClaw framework.
Q&A Highlights
13:15 — Can VSS run fully offline at the edge?
Yes, supported on AGX, IGX, and DGX Spark. Broader edge support coming in v3.2.
10:48 — Is VSS free to use?
Yes, free and open source. Developer license needed for NGC containers. Full source code on GitHub June 1st.
54:10 — Can we deploy multiple VSS profiles at once?
Yes, but they should share VLM endpoints to avoid resource overlap. Easier multi-profile support coming in June.
58:27 — Can an agent create custom skills from plain English?
Yes — chat through a workflow with the agent, then ask it to package it into a reusable skill..
0:00–21:35 VSS Overview, Architecture, Skills & NemoClaw
21:36–23:25 Demo Part 1 – Deploying VSS with Codex
23:26–24:39 Setting Up VSS on Brev Launchable
24:40–31:24 VSS GitHub Skills Walkthrough
31:25–35:17 Installing VSS Skills with Codex
35:18–42:02 Deploying the VSS Search Profile & OpenClaw Setup
42:03–44:18 Demo Part 2 – Run video analytics AI agent with OpenClaw
44:19–46:39 Semantic Fusion Video Search
46:40–49:06 Visual Verification with Cosmos Reason 2
49:07–51:39 AI-Generated Analytics Dashboard
Resources
📚VSS Build: nvda.ws/4wnM71T
📚VSS Skills: nvda.ws/4ts8xMK
📚VSS Doc: nvda.ws/4nnuj2M
📚VSS Tech blog - nvda.ws/4d7RrPE
Organizations increasingly rely on video to capture critical information—yet extracting meaningful insights from massive amounts of footage remains a challenge. NVIDIA Metropolis Blueprint for video search and summarization (VSS) overcomes this hurdle by transforming hours of video into instantly searchable, actionable intelligence.
With advanced indexing, scalable architecture, and real-time analytics, VSS helps enterprises monitor operations, detect trends, and make informed decisions faster than ever. VSS redefines video search, analysis, and reporting through next-generation multimodal intelligence and modular design.
Now with NVIDIA NemoClaw, an open source stack that adds privacy and security controls to OpenClaw, you can deploy video agents using VSS skills.
What you'll learn:
Build a video agent with VSS skills on Brev launchable and deploy a search profile for semantic fusion search with NemoClaw.
Use VSS skills with NemoClaw to integrate agents into Slack to search videos and review clips from chat.
Configure agents to generate real-time VLM-powered alerts from a live streaming video.
Understand the end‑to‑end pipeline to build agents and monitoring workflows with the OpenClaw framework.
Q&A Highlights
13:15 — Can VSS run fully offline at the edge?
Yes, supported on AGX, IGX, and DGX Spark. Broader edge support coming in v3.2.
10:48 — Is VSS free to use?
Yes, free and open source. Developer license needed for NGC containers. Full source code on GitHub June 1st.
54:10 — Can we deploy multiple VSS profiles at once?
Yes, but they should share VLM endpoints to avoid resource overlap. Easier multi-profile support coming in June.
58:27 — Can an agent create custom skills from plain English?
Yes — chat through a workflow with the agent, then ask it to package it into a reusable skill..
0:00–21:35 VSS Overview, Architecture, Skills & NemoClaw
21:36–23:25 Demo Part 1 – Deploying VSS with Codex
23:26–24:39 Setting Up VSS on Brev Launchable
24:40–31:24 VSS GitHub Skills Walkthrough
31:25–35:17 Installing VSS Skills with Codex
35:18–42:02 Deploying the VSS Search Profile & OpenClaw Setup
42:03–44:18 Demo Part 2 – Run video analytics AI agent with OpenClaw
44:19–46:39 Semantic Fusion Video Search
46:40–49:06 Visual Verification with Cosmos Reason 2
49:07–51:39 AI-Generated Analytics Dashboard
Resources
📚VSS Build: nvda.ws/4wnM71T
📚VSS Skills: nvda.ws/4ts8xMK
📚VSS Doc: nvda.ws/4nnuj2M
📚VSS Tech blog - nvda.ws/4d7RrPE
