The physics layer
for physical AI.

Turn any 3D mesh into a simulation-ready environment in seconds. No manual tuning. No guesswork. Built for MuJoCo, Gazebo, PyBullet, and Isaac Sim.

Used by robotics teams training on MuJoCo · Gazebo · PyBullet

0.00s

Average

processing time

0

Physics profiles

per scene

0

DR ranges

computed automatically

Simulation setup is still a 3-month manual nightmare.

[ Manual Pipeline ]

Rebuild collision meshes manually

Estimate mass and density (guesswork)

Tune friction via trial-and-error

Maintain simulator-specific XML configs

Block training for 3+ months

[ EnviScale API ]

Upload raw .obj / .glb mesh

Infer mass, density, friction automatically

Generate domain-randomized surface profiles

Export MuJoCo, Gazebo, PyBullet formats

Simulation-ready in <5 seconds

Manual Pipeline

Week 1
Week 4
Week 8
Wk 12 → Run

EnviScale API

Upload
5s → First run ✓

Stop tuning physics. Start training robots.

No pipeline. No guesswork. Just a compiler for physics.

Upload your mesh

.obj .ply .glb .gltf .stl

Drag & drop or REST API call

mesh validated

mug.glb · 18,240 vertices · watertight

Simulation parameters
computed from geometry

> material_classifierceramic (0.84)
> density_lookup2400 kg/m³
> mass_solver0.31 kg
> friction_modelcomputed

Computed in 2.09s · Deterministic · Simulator-ready

Download your
simulation files

scene_clean.xmlbaseline physics
scene_worn.xmlreduced friction
scene_contaminated.xmlworst-case conditions

+ physics_report.json

Full confidence report, DR training ranges,
collision hull metadata

MuJoCo · Gazebo · PyBullet ready

From mesh to simulation in seconds.

Not a pipeline. A compiler for physics.

  • Mass, friction, and inertia derived from geometry
  • Deterministic outputs for reproducible training
  • Simulation-ready across engines instantly

Same object. Different surface conditions.

clean
worn
contaminated

PROFILE: [clean]

01/03

Perfect lab conditions • high friction • stable contact

Mass0.310 kg
Friction (slide)0.600
Friction (torsion)0.005
Restitution0.200
Confidence84%

DR TRAINING RANGE

mass:[0.23 → 0.40] kg
friction:[0.42 → 0.81]

scene_clean.xml ready to load

PROFILE: [worn]

02/03

After repeated use • smoother surfaces • reduced grip

Mass0.310 kg
Friction (slide)0.360
Friction (torsion)0.003
Restitution0.200
Confidence84%

DR TRAINING RANGE

mass:[0.23 → 0.40] kg
friction:[0.25 → 0.49]

scene_worn.xml ready to load

PROFILE: [contaminated]

03/03

Dust or oil present • low friction • worst-case handling

Mass0.310 kg
Friction (slide)0.210
Friction (torsion)0.002
Restitution0.200
Confidence84%

DR TRAINING RANGE

mass:[0.23 → 0.40] kg
friction:[0.15 → 0.28]

scene_contaminated.xml ready to load

One API call. Complete simulation package.

No SDK. No complex setup. Just curl.

request.sh
 
 
 
 
 
 
 
 
 
 
 
 
 
 
api.enviscale.com
 
 
 
 
 
 

Simple pricing. No simulation PhD required.

MonthlyAnnual (save 20%)

Traditional manual setup: ~$10k/scene. EnviScale: $299/mo.

FREE

$0

3 scenes/month

MuJoCo export
3 surface profiles
Physics report
DR ranges
API access
Get started free
Most Popular

STARTER

$299/month

50 scenes/month

All FREE features
Gazebo + PyBullet export
Fill state profiles
DR ranges
REST API + key
Email support
Start free trial

PRO

$799/month

Unlimited scenes

All STARTER features
All simulator formats
Priority API access
DR ranges
Dedicated Slack channel
Custom integrations
Contact us

The physics layer

for physical AI.

Built for robotics teams training on MuJoCo,Gazebo, and Isaac Sim.

Early access · limited spots available.