NAB 2026: NVIDIA’s Not-So-Secret AI Agents
One demo I saw at NAB 2026 covered using agents to create content. Obviously, agents need to be managed so they don't think too far outside the box. In this demo, NVIDIA talked about their control plane for a multitask agent project that helps create both a script and animated characters.
Agent Background
“Thinking about agents, this is really the fundamental shift from retrieval-based computing, where you create a file, access, store, work on it some more and save it,” said NVIDIA Director Global Media and Entertainment Strategy and Marketing Rick Champagne. “We’re moving [to] AI factories where everything is generated and iterated upon.”
Champagne went on to explain that “agents are using reasoning and planning and solving multi-step problems. They can act with the autonomy that you allow.”
NVIDIA has a new framework for managing the multitude of agents you may need to execute a task with enterprise-grade guardrails. They recently introduced NemoClaw for OpenClaw, an open-source stack that adds privacy and security controls to OpenClaw, the company’s open-source tool for running agents and was recently acquired by OpenAI.
NemoClaw enforces policy-based privacy and gives users control over how agents behave and handle data. “From a trust standpoint, you also have to control identity and authentication,” Champagne said. “That’s the agent’s core observability, and you want to be able to see what your agents are doing. You want to audit their behavior; that’s really critical as agents become more and more autonomous,” he explained.
OpenClaw also evaluates available compute resources to run high-performance open models like NVIDIA Nemotron locally for enhanced privacy and cost efficiency.
For developers building AI-powered projects, this removes one of the biggest blockers to deploying autonomous agents in production. The idea, according to one developer, is that this will help you avoid shocking monthly bills for cloud hosting, API costs, and proprietary platforms.
The same developer’s Linkedin post says they can deploy autonomous AI agent with a single terminal command and what they like are:
- Built-in guardrails prevent your agent from going off-script
- Open-source MIT License
- No proprietary lock-in, works with any model
- Runs 24/7 locally or in the cloud via NVIDIA Brev, a launchable free GPU environment
- Partners like Adobe, Salesforce, and SAP
Demo
Champagne told the NAB audience that NVIDIA has recently released NVIDIA AI for media, a suite of audio, video, and augmented reality SDKs. The company also provides inference microservices. “You can string them together and you can build your own agents," Champagne said.
The following use cases are listed on NVIDIA’s website:
- Content localisation blueprint
- Synthetic Video detector
- LipSync
- Active speaker detection
- Background noise removal
- Studio voice NIM
- Video relighting
- RTX video super resolution
- 3D body pose
- Audio effects
- Video effects
- Augemented reality
One example is taking out background noise from narration and then upscaling the recording. Another is localising content archives to additional languages. The demo showed video of a character talking in a perfectly synced video clip in 8 different languages.
The demo also showed off creative agents which were able to work with visuals to combine two different objects, like a black fox and an insect. The artist is then able to further refine the imagery, changing the content based on a reference image of a particular style.
The demo also showed how the agents can help creatives pitch story ideas. Taking a story written by an NVIDIA employee, the AI “reformats it, edits the log line for the story, generates shot breakdowns, and does character profiles."
All of this is editable, like any prompt system. If you don’t like what it produces, you can further elaborate on what changes you want to see. “It’s gone through the script, it’s found the characters, and it’s automatically generated some images for you,” said Champagne. “The whole idea here is we’re just pitching the story. It does that automatically, and then it creates a character profile sheet."
The demo went on to walk viewers through the next step, shot breakdowns: “going through the script and it starts to create a shot breakdown. So, it’s creating some images... pulling characters together, applying a style, and then from there you can go and you can fine tune.”
Both the model and the user are learning from the prompt interaction. You can also create a 3D gaussian splat of a still photo, so you can reimagine it from different angles and fill in missing imagery within the picture to create a better model. If you're not up for building things yourself, system integrators from NVIDIA are the way to go.
Related Articles
My visit to NAB 2026 skewed heavily toward talking about workflow changes with AI. I tried to keep my travels on the convention floor to very specific examples of AI's impact on streaming workflows, and this article will explore the most interesting ones I found. First up is how AMC Global Media (newly rebranded from AMC Networks) is working with AWS to leverage AI to create better content access.
27 Apr 2026
Perhaps the booths that cropped up on my NAB 2026 itinerary and the sessions I squeezed in between simply skewed this way, but it seems like everyone's favorite sport in streaming this year is, well, sport. Certainly, when it comes to live, the pinnacle of achievement appears to be seamless delivery of premium sports events at global scale, with an eye to "meeting fans exactly where they are"—crossing national, lingual, and cultural borders via localisation, and transcending generations through autogenerated highlights and verticalised mobile delivery. And when it comes to traditional landscape sport video streamed to conventional CTV glass, multiview more than ever seems to be the name of the game in 2026.
21 Apr 2026
As we head into another NAB, I'm less interested in what looks impressive on a stand and more interested in what is actually ready to survive in production.
15 Apr 2026