This product was not featured by Product Hunt yet. It will not be visible on their landing page and won't be ranked (cannot win product of the day regardless of upvotes).
Rigyd
Simulation-ready 3D assets for robotics simulation, at scale
Robot learning is only as good as its data. For sim-first teams, success depends on domain randomization. Train a humanoid in 1M simulated kitchens so it doesn't fail in the real one. Or train a robotic arm on thousands of objects of varying sizes, masses, and materials. Both require physics-accurate 3D content at scale. Most teams don't have it. Rigyd auto-generates and randomizes physics-accurate 3D objects and worlds at scale, unlocking sim-to-real scenarios that were previously impossible.
👋 Hey Product Hunt, I'm Ugur, co-founder of Rigyd.
For the past 4 years, we've built a platform that generates 3D visuals and powers AR try-on experiences for some of the largest footwear and retail brands in the world. We've processed 30K+ 3D assets and delivered immersive experiences to 10M+ users in the last 12 months at artlabs.ai. The goal was to give shoppers a photorealistic look, so we obsessed over mesh fidelity and texture quality.
Then robotics companies started knocking. They wanted millions of 3D assets to populate simulation environments. Sim-to-real transfer works through domain randomization: train your humanoid in 1M different kitchens and a real one isn't a surprise when it's deployed.
That's when we realized that looking right and behaving right are completely different problems. These teams needed proper collision meshes, physically accurate mass, realistic friction, and more. Not just pretty geometry and texture.
Most 3D assets weren't built for physics. A robot trained on them learns nothing useful. It falls through floors, grips through objects, and generalizes terribly to the real world.
And it's not the simulators holding things back. Isaac Sim, MuJoCo, and Gazebo are widely deployed. Newer engines like Genesis have arrived with even faster runtimes. The physics engines are mature. The 3D content going into them isn't.
The current fixes fall short. Teams either manually create or annotate every asset (doesn't scale) or accept platform defaults and hope the policy generalizes. Neither works.
So we built Rigyd. It ingests raw 3D, images, or text and produces validated OpenUSD (with USDPhysics schemas) and MJCF, with collision meshes, mass, friction, restitution, material properties, and semantic labels baked in. Works in Isaac Sim, MuJoCo, and Unreal out of the box.
No manual annotation. Weeks → minutes. From any input.
Our early users are already using it for → Warehouse automation → Simulating surgical rooms → Testing robotic arms before deployment
The UI is for exploration. The real surface is programmatic. Rigyd exposes tools so agents (or your own pipelines) can generate millions of physics-accurate sims directly inside NVIDIA Isaac Sim, MuJoCo, or wherever you train.
If you're building anything in robotics, AV, industrial automation, or embodied AI: claim the free credits on signup. All we'd ask is honest feedback. What works, what doesn't, what you wish it did differently.
If you've hit the sim-data wall and want enterprise/bulk API access, DM me. We're onboarding a small group of design partners in June.
If you just want to nerd out about OpenUSD, domain randomization, or sim-to-real transfer, we're in the comments all day.
Robotics is going to be the largest industry in history. We think the data infrastructure for it should be as easy to use as Stripe is for payments. That's what we're building.
Congrats on the launch. Also love the OpenUSD angle. I believe like the ecosystem is converging around more standardized pipelines instead of every company building tooling around meshes && annotations :)
Do you randomize visuals too, or is Rigyd mainly focused on physics and geometry?
One of the biggest bottlenecks in robotics today is not the model, it’s the data. Rigyd tackles a massive problem by making physics-accurate synthetic training environments scalable and practical.
Training across millions of variations is exactly what humanoid and robotic manipulation systems need to generalize in the real world. Excited to see where this goes.
This is cool. Curious how are you generating collision meshes under the hood? VHACD-style decomposition, SDFs, or custom approximations?
Wow, quality of the generated assets are impressive. With all the materials and physics. 3 credits for trial is not enough though to test all the features.
Congrats on the launch! Looks great. Curious what asset formats you ingest (OBJ, FBX, GLB, STEP, USD)?
How do you validate that the physics actually behaves correctly post-generation? Do you run automated sim rollouts, compare against ground-truth measurements, or rely on schema conformance only?
About Rigyd on Product Hunt
“Simulation-ready 3D assets for robotics simulation, at scale”
Rigyd was submitted on Product Hunt and earned 57 upvotes and 25 comments, placing #20 on the daily leaderboard. Robot learning is only as good as its data. For sim-first teams, success depends on domain randomization. Train a humanoid in 1M simulated kitchens so it doesn't fail in the real one. Or train a robotic arm on thousands of objects of varying sizes, masses, and materials. Both require physics-accurate 3D content at scale. Most teams don't have it. Rigyd auto-generates and randomizes physics-accurate 3D objects and worlds at scale, unlocking sim-to-real scenarios that were previously impossible.
Rigyd was featured in Robots (10.6k followers), Artificial Intelligence (469k followers) and 3D Modeling (2k followers) on Product Hunt. Together, these topics include over 99.9k products, making this a competitive space to launch in.
Who hunted Rigyd?
Rigyd was hunted by Ugur Yekta Basak. A “hunter” on Product Hunt is the community member who submits a product to the platform — uploading the images, the link, and tagging the makers behind it. Hunters typically write the first comment explaining why a product is worth attention, and their followers are notified the moment they post. Around 79% of featured launches on Product Hunt are self-hunted by their makers, but a well-known hunter still acts as a signal of quality to the rest of the community. See the full all-time top hunters leaderboard to discover who is shaping the Product Hunt ecosystem.
Want to see how Rigyd stacked up against nearby launches in real time? Check out the live launch dashboard for upvote speed charts, proximity comparisons, and more analytics.
Ugur Yekta Basak
👋 Hey Product Hunt, I'm Ugur, co-founder of Rigyd.
For the past 4 years, we've built a platform that generates 3D visuals and powers AR try-on experiences for some of the largest footwear and retail brands in the world. We've processed 30K+ 3D assets and delivered immersive experiences to 10M+ users in the last 12 months at artlabs.ai. The goal was to give shoppers a photorealistic look, so we obsessed over mesh fidelity and texture quality.
Then robotics companies started knocking. They wanted millions of 3D assets to populate simulation environments. Sim-to-real transfer works through domain randomization: train your humanoid in 1M different kitchens and a real one isn't a surprise when it's deployed.
That's when we realized that looking right and behaving right are completely different problems. These teams needed proper collision meshes, physically accurate mass, realistic friction, and more. Not just pretty geometry and texture.
Most 3D assets weren't built for physics. A robot trained on them learns nothing useful. It falls through floors, grips through objects, and generalizes terribly to the real world.
And it's not the simulators holding things back. Isaac Sim, MuJoCo, and Gazebo are widely deployed. Newer engines like Genesis have arrived with even faster runtimes. The physics engines are mature. The 3D content going into them isn't.
The current fixes fall short. Teams either manually create or annotate every asset (doesn't scale) or accept platform defaults and hope the policy generalizes. Neither works.
So we built Rigyd. It ingests raw 3D, images, or text and produces validated OpenUSD (with USDPhysics schemas) and MJCF, with collision meshes, mass, friction, restitution, material properties, and semantic labels baked in. Works in Isaac Sim, MuJoCo, and Unreal out of the box.
No manual annotation. Weeks → minutes. From any input.
Our early users are already using it for
→ Warehouse automation
→ Simulating surgical rooms
→ Testing robotic arms before deployment
The UI is for exploration. The real surface is programmatic. Rigyd exposes tools so agents (or your own pipelines) can generate millions of physics-accurate sims directly inside NVIDIA Isaac Sim, MuJoCo, or wherever you train.
If you're building anything in robotics, AV, industrial automation, or embodied AI: claim the free credits on signup. All we'd ask is honest feedback. What works, what doesn't, what you wish it did differently.
If you've hit the sim-data wall and want enterprise/bulk API access, DM me. We're onboarding a small group of design partners in June.
If you just want to nerd out about OpenUSD, domain randomization, or sim-to-real transfer, we're in the comments all day.
Robotics is going to be the largest industry in history. We think the data infrastructure for it should be as easy to use as Stripe is for payments. That's what we're building.
Check it out 👉 rigyd.com
Ugur