Uploaded June 2026 | Updated September 2026, 2 weeks ago
In this tutorial, you will learn how to use NVIDIA physical AI agent skills and tools to generate multiple realistic variations of a single robot mobility clip using a simple natural language prompt. The Video Augmentation Agent Skill runs in a preconfigured NVIDIA Brev launchable, letting you change lighting, color, environment conditions, and visual style across a video dataset without manually stitching together fragmented tools. By the end, you will have the video augmentation Agent Skill running end-to-end, from generation through evaluation, with the agent ranking outputs by realism and visual diversity automatically.
Get started with the Video Augmentation Agent Skill:s:
🚀 Deploy the NVIDIA Brev Launchable → brev.nvidia.com/launchable/deploy?launchableID=env-3DH5QPdLRWBABdW9Lxn4vQBK5mW
⬇️ Download the open-source skill on GitHub → github.com/NVIDIA/skills/tree/main/skills/physical-ai-video-data-augmentation
Chapters:
00:00 Overview of available NVIDIA Physical AI Launchables
00:15 Intro — what the Video Data Augmentation launchable produces
00:25 Preview of augmented output variations (robotics warehouse clips)
00:55 Preview of augmented output variations (autonomous driving clips)
01:15 Launchable overview page — pipeline features walkthrough
01:35 Deploy the Video Data Augmentation launchable on brev.nvidia.com
01:40 Instance ready — view secure links and open OpenClaw
02:00 Send augmentation prompt for 6 warehouse video variations
02:30 Agent locates source video and identifies the correct skill
02:55 Agent submits augmentation jobs and tracks progress per variant
03:20 Check workflow status in OSMO
03:45 Agent surfaces completed videos — preview warehouse_contrast.mp4
04:05 Ask agent to evaluate and rank all 6 videos for model training utility
04:05 Final output — 6 augmented warehouse variations shown side by side
#NVIDIAAgentSkills #PhysicalAI #SyntheticData
NVIDIA Agent Skills, video augmentation, synthetic data generation, robot simulation, Physical AI, NVIDIA Launchable
Q: What is the NVIDIA video augmentation agent skill?
A: It is a launchable workflow that takes a robot simulation video and generates multiple realistic variations — changing lighting, color, environment conditions, and visual style — using a simple chat prompt. The agent handles generation, status tracking, and evaluation in a single end-to-end pipeline.
Q: What kinds of variations can the video augmentation agent skill generate?
A: In this walkthrough the agent generates six augmented versions of a single clip, including variations in contrast, lighting, and visual style. You can specify the types of augmentations you want using plain language prompts.
Q: What credentials do I need to get started?
A: You need a Hugging Face token, an NGC API key, and an inference API key from build.nvidia.com. These are set up once during the OpenCLAW credential configuration step before you start generating.
Q: What is OpenCLAW and how does it relate to the agent skill?
A: OpenCLAW is the interactive chat UI that comes preconfigured with all the skills needed to complete the video augmentation workflow end-to-end. You interact with it through plain language prompts to trigger generation, check status, and evaluate outputs.
Q: What is OSMO and when should I use it?
A: OSMO is the backend orchestration tool the agent uses to manage tasks. It shows a history of all your augmentation runs, real-time progress for each video, and logs useful for debugging. You can check it any time during or after a generation run.
In this tutorial, you will learn how to use NVIDIA physical AI agent skills and tools to generate multiple realistic variations of a single robot mobility clip using a simple natural language prompt. The Video Augmentation Agent Skill runs in a preconfigured NVIDIA Brev launchable, letting you change lighting, color, environment conditions, and visual style across a video dataset without manually stitching together fragmented tools. By the end, you will have the video augmentation Agent Skill running end-to-end, from generation through evaluation, with the agent ranking outputs by realism and visual diversity automatically.
Get started with the Video Augmentation Agent Skill:s:
🚀 Deploy the NVIDIA Brev Launchable → brev.nvidia.com/launchable/deploy?launchableID=env-3DH5QPdLRWBABdW9Lxn4vQBK5mW
⬇️ Download the open-source skill on GitHub → github.com/NVIDIA/skills/tree/main/skills/physical-ai-video-data-augmentation
Chapters:
00:00 Overview of available NVIDIA Physical AI Launchables
00:15 Intro — what the Video Data Augmentation launchable produces
00:25 Preview of augmented output variations (robotics warehouse clips)
00:55 Preview of augmented output variations (autonomous driving clips)
01:15 Launchable overview page — pipeline features walkthrough
01:35 Deploy the Video Data Augmentation launchable on brev.nvidia.com
01:40 Instance ready — view secure links and open OpenClaw
02:00 Send augmentation prompt for 6 warehouse video variations
02:30 Agent locates source video and identifies the correct skill
02:55 Agent submits augmentation jobs and tracks progress per variant
03:20 Check workflow status in OSMO
03:45 Agent surfaces completed videos — preview warehouse_contrast.mp4
04:05 Ask agent to evaluate and rank all 6 videos for model training utility
04:05 Final output — 6 augmented warehouse variations shown side by side
#NVIDIAAgentSkills #PhysicalAI #SyntheticData
NVIDIA Agent Skills, video augmentation, synthetic data generation, robot simulation, Physical AI, NVIDIA Launchable
Q: What is the NVIDIA video augmentation agent skill?
A: It is a launchable workflow that takes a robot simulation video and generates multiple realistic variations — changing lighting, color, environment conditions, and visual style — using a simple chat prompt. The agent handles generation, status tracking, and evaluation in a single end-to-end pipeline.
Q: What kinds of variations can the video augmentation agent skill generate?
A: In this walkthrough the agent generates six augmented versions of a single clip, including variations in contrast, lighting, and visual style. You can specify the types of augmentations you want using plain language prompts.
Q: What credentials do I need to get started?
A: You need a Hugging Face token, an NGC API key, and an inference API key from build.nvidia.com. These are set up once during the OpenCLAW credential configuration step before you start generating.
Q: What is OpenCLAW and how does it relate to the agent skill?
A: OpenCLAW is the interactive chat UI that comes preconfigured with all the skills needed to complete the video augmentation workflow end-to-end. You interact with it through plain language prompts to trigger generation, check status, and evaluate outputs.
Q: What is OSMO and when should I use it?
A: OSMO is the backend orchestration tool the agent uses to manage tasks. It shows a history of all your augmentation runs, real-time progress for each video, and logs useful for debugging. You can check it any time during or after a generation run.
