Wano StudioComing soon

Synthetic data
that doesn't cheat.

Teach your robot. Simply.

You describe. Your robot learns.

Wano Studio is the workshop that prepares your robot's training. Describe a task once: Wano builds the scenes, places the robot and spots the shortcuts before training even starts.

3D mock-up of the task: a white robot arm carries a blue cube toward a green goal zone. Its path is drawn as dashed ghost cubes. In red, the shortcut: the cube held above the zone, never put down. After the fix, the red path goes out and a blue path leads the cube down into the zone.
Goal zoneShortcutFixed
Anti-cheatAnd 9 good checks.Attention

Reaching the area is enough: nothing asks to release the object or for it to rest.

If you leave this, the robot may hold the object above the area without putting it down.

FixAdd “released and at rest at the end”On the mock-up

Works with the stack you already use

  • Isaac Lab
  • MuJoCoPhysics simulation
  • Robot simulation
  • 2Real-world deployment
  • Datasets & training

Synthetic data

Real gravity. Real friction.

Wano Studio creates synthetic data: training demonstrations produced in simulation, instead of recorded one by one on a real robot. And every one of them obeys the laws of physics.

What is synthetic data?

Trials generated by a computer to train a robot. Thousands of demonstrations, without weeks of hands-on recording or broken hardware. But not all of it is equal.

Stuck

Made for the picture

The glass sticks to the gripper.

Some synthetic data is made to look good: the object “sticks” as soon as the gripper closes. No friction, no slipping, no weight. The video looks great; the robot learns nothing real.

SlippingHeld

With Wano

The glass can slip.

Wano really simulates. If the grip is too weak, the glass slips and drops back. The robot learns to squeeze just enough, as in the real world.

What Wano really simulates

  • Gravity

    What you let go of falls. What you carry weighs.

  • Friction

    A grip holds thanks to friction, not glue. Too weak, and it slips.

  • Mass

    A heavy object takes more force, and takes longer to stop.

  • The robot's limits

    What it can carry, and how far out: the strength of its motors counts.

  • Joints

    Every joint keeps its real limits. No wrist spinning forever.

  • Contacts

    Objects touch, push and block each other. Nothing passes through anything.

Why it matters

A robot that has only known sticky objects fails at its first real glass. Physically accurate data narrows the gap between simulation and the real world (the sim-to-real gap): what your robot learns in Wano, it learns for real.

How it learns · 4 chapters

One small task. A world of learning.

Follow one cube from your description to the right gesture. Each chapter shows what Wano Studio does for you, live.

The task we follow

“Put the cube in the area.”

Same task. Every chapter.

01 · Describe

You describe it once.

Answer a few questions: which object, where it starts, where it should go, which robot. A white mock-up of the task builds itself as you answer.

You give
A few answers
You get
The task, as a 3D mock-up
A white 3D mock-up builds itself: a table, a blue cube, a dashed start area, a blue goal area, then the robot arm.

01Describe

The tableThe cubeWhere it startsWhere it goesThe robot

Wano Studio · Step by step

From your first idea to a learned gesture.

Compose the task, check the scenes, generate demonstrations and follow training. One workflow, inside the Studio.

Explore the workflow
  1. Animated preview · 4 sGIF

    01The catalog

    Choose the objects. Keep your options.

    A table, a block, a plate. Select their types and browse available models. Your template can reuse different objects from the catalog.

    One choice of types, many possible scenes.

  2. Animated preview · 4 sGIF

    02Placement

    Same gesture. A new starting point.

    At the front, on the left, or anywhere on the surface: choose a starting area and allowed rotations. The plan immediately shows your selection.

    Vary positions within clear boundaries.

  3. Animated preview · 4 sGIF

    03The goal

    Say where the object should land.

    “Place the block on the plate.” Connect the instruction to a destination in the scene. The preview makes the expected result concrete.

    A clear instruction, a visible destination.

  4. Animated preview · 4 sGIF

    04Cameras

    Choose where the robot sees from.

    Choose a reference and vary positions between demonstrations. Switch from perspective to top view to understand the camera layout.

    Visual diversity is a setting you control.

  5. Animated preview · 4 sGIF

    05Lighting

    A new morning. A different light. The same goal.

    Select lighting conditions: morning, daytime, overcast or dusk. The generator varies the light between demonstrations.

    More visual variety to cover with your examples.

  6. The uFactory Lite 6 arm, rendered in white: its wrist ends in a bare flange, with no gripper.uFactory Lite 6

    No gripper: it can't pick the object up

    The SO-ARM100 arm, rendered in white, with its two-jaw gripper open.SO-ARM100

    Has a gripper, and reaches the object

    Product preview

    06The robot

    The right robot, within reach.

    Wano checks whether the robot fits the task, then places it within reach of the object in each scene. A model without a gripper cannot perform a grasping task.

    Choose a robot that can perform the gesture.

  7. 3D mock-up of the task: a white robot arm carries a blue cube toward a green goal zone. Its path is drawn as dashed ghost cubes. In red, the shortcut: the cube held above the zone, never put down. After the fix, the red path goes out and a blue path leads the cube down into the zone.
    Goal zoneShortcutFixed

    Reach the target, release the object, let it settle.

    Product preview

    07Verification

    Catch mistakes before training.

    Reaching the target while still holding the object is a shortcut, not success. Wano shows the risk on the model and checks reach, grasping and rules. A blocking check prevents training from starting.

    Understand the problem, then correct the rule.

  8. Animated preview · 4 sGIF

    08Demonstrations

    Synthetic teleoperation.

    Let the agent automatically generate demonstrations across thousands of varied scenes. Review the resulting motions and the examples kept for training. You can also guide the robot yourself through manual teleoperation. Inspect each motion in mosaic view or one camera at a time.

    Automate demonstrations, or take control yourself.

  9. Data quality

    • Episode recorded in a scene drawn from the task.Episode 1
    • Episode recorded in a scene drawn from the task.Episode 2
    • Episode recorded in a scene drawn from the task.Episode 3Hesitant
    • Episode recorded in a scene drawn from the task.Episode 4
    • Episode recorded in a scene drawn from the task.Episode 5
    • Episode recorded in a scene drawn from the task.Episode 6
    • Episode recorded in a scene drawn from the task.Episode 7
    • Episode recorded in a scene drawn from the task.Episode 8Jerky
    • Episode recorded in a scene drawn from the task.Episode 9
    • Episode recorded in a scene drawn from the task.Episode 10
    • Episode recorded in a scene drawn from the task.Episode 11Too long
    • Episode recorded in a scene drawn from the task.Episode 12

    Nothing is deleted.

    Product preview

    09Data quality

    Keep examples that teach the right gesture.

    Spot hesitant, jerky or overly long episodes. Set them aside to prepare training; you can undo that choice. Try the preview.

    Nothing is deleted.

  10. Animated preview · 4 sGIF

    10Learning

    Follow progress, not just a waiting screen.

    Training steps and the loss curve make the run readable. Return to demonstrations through the stages of the same workflow.

    Imitation loss and task success measure different things.

  11. Illustrative exampleGIF

    11Diagnosis

    See what still needs learning.

    Compare familiar scenes with held-out test scenes. This illustrative example shows 80% success on familiar scenes and 70% on test scenes. The report helps you choose what to improve next.

    A finished run does not validate the behavior.

Product captures · Explanatory zooms · Interface shown in French. Diagnosis scores are illustrative examples, not measured results or a performance promise.

Two films · 35 seconds each

Your robot hasn't moved yet. Wano already sees the shortcut.

First, Wano Studio catches a shortcut on the mock-up, before any training. Then it says the risk in one sentence, and fixes it in one click.

1 · The shortcut

Wano sees it before training

2 · The fix

One risk. One sentence. One click.

Questions

Good questions. Straight answers.

What people ask before they try Wano Studio.

01Do I need to know how to code?

No. You answer a few simple questions about the task. Wano builds the mock-up, the scenes and the checks for you.

02Is your synthetic data generated video?

No. Every trial is simulated with MuJoCo: gravity, friction, mass and your robot's real limits. If an object should slip, it slips.

03Do I need a big GPU or a robot lab?

No. Wano Studio runs on your own computer, and a Mac is enough to start. Every trial happens in simulation first, with MuJoCo.

04Does my data leave my machine?

No. Your tasks, scenes, trials and data are plain files on your computer. Nothing is sent anywhere unless you choose to.

05Which robots can I use?

Arms with a gripper, like the SO-ARM100, are ready to go, and you can bring your own. Either way, Wano checks the robot can do your task before you pick it.

06When can I try it?

Wano Studio is opening to its first users, in small groups. Join the waitlist and we'll let you know.

Waitlist

Check before you train.

Wano Studio is opening to its first users. Leave your email and we'll let you know.