Today, we’re launching a new capability that connects everything we’ve been building directly to the robotics world: free USDZ export. Turn any real-world space into a robot training ground. For FREE. Capture with your smartphone or a 360 cam → get a simulation-ready USDZ (Gaussian splat + aligned mesh)... → drop it straight into NVIDIA Isaac Sim. No LiDAR rig. No sales call. 🧵👇show more

2/ Every capture becomes: ☁️ Gaussian splat = the robot's eyes. Photoreal RGB from any viewpoint, like its real cameras 📐 Aligned mesh = the physics. Contact, navigation, manipulation 📏 Metric scale: 1 m = 1 m One USDZ: interact physically, perceive photorealistically.

3/ The sim-to-real gap is real. Policies trained in synthetic scenes stumble in messy reality. A real-world splat carries what synthetic can't fake: true lighting, textures, clutter. Train in a metric-accurate twin of the site → your robot shows up already knowing the place.

4/ The economics ❌ Traditional RGB-LiDAR rigs: tens of thousands of $ ✅ A 360 cam: ~$500 ✅ The phone in your pocket: $0 extra Walk a street or a warehouse once and it's captured. Environment-specific training data just stopped being a luxury.

5/ How it works 1️⃣ Capture with smartphone or 360 cam 2️⃣ Upload to OVER 3️⃣ We generate the splat + aligned mesh, packaged as one USDZ 4️⃣ Import into Isaac Sim / Isaac Lab and start training Real-to-sim in an afternoon. $0. cc @NVIDIARobotics @nvidiaomniverse

6/ To our community: this is what map2earn™ was always building toward. 🌍 Every scene you capture isn't just AR real estate, it's training infrastructure for Physical AI. Mappers earn. Robots learn. DePIN meets Physical AI. ethereum:0x21bfbda47a0b4b5b1248c767ee49f7caa9b23697

7/ We didn't start yesterday: the OVER community has already produced 270K+ mappings, 105M+ images of real-world places. The most valuable dataset for Physical AI won't be built by one company with LiDAR vans. It'll be built by everyone, phone in hand.

8/ Read the full blog post: Try it now 🔗 Capture your world → download your USDZ → train your robot. Reality is the ultimate training set. Now it's free to use. 🤖🌍

@1NeverQ

Making real-world data easier to use is a win

This could speed up AI training significantly

Making robot training more accessible. Great update

bravi ragazzi continuate a costruire💪

This makes experimenting with robotics a lot more practical.

This help removes another barrier to robotics development

Love seeing the gap between the physical and virtual worlds getting smaller. Excited to see what builders create with this. 🚀

This removes a huge barrier for robotics builders.

This removes a huge barrier for robotics developers. Capturing a real-world environment and taking it straight into Isaac Sim—for free—is a big step forward. 🤖

Massive boots for this one, great stuffs for real

This is a great step for physical AI

omo this is huge for sim-to-real pipelines

Big win for robotics builders. Excited to see this in action

This is the kind of progress the space needs. 🔥

This makes robotics development way more accessible. Big move.

No LiDAR rig needed for robot training data is a huge barrier removed.

Love seeing tools become easier to use

The phone-to-simulation workflow is probably my favorite part of this.

This is a huge win for builders. 🚀

this is the kind of ai infra that quietly changes everything

Love seeing tools become easier to use

Big step for Physical AI

This makes robotics development much more accessible

Turning real spaces into training environments is impressive

Excited to see what builders create with this

Nice, keep building 🤝

Turning a phone into a simulation tool is a big win.