Decentralized render network

Every idle GPU
becomes your
render farm

Submit a scene. The network slices it into tiles, renders them across consumer GPUs worldwide, and proves every frame it hands back. No cluster to provision. No queue to sit in.

Submit a job Run a node
Nodes online—
Frames delivered—
Aggregate— TF
Render frontier Tiles —● live
Pass 1 Sample 64 Verified 100% Job —
01Model
What it is

Rendering as a
utility, not a purchase

A render farm is a terrible thing to own: idle most of the year, obsolete in three, and you pay for all of it. LUMEN turns the same work into a metered utility drawn from GPUs that were going to be idle anyway.

A wireframe resolving into a solid form
01 / SUBMIT

Submit

Send a scene or a job spec with a quality tier and a ceiling on spend. Tiles are handed out in seconds — not queued behind another studio's deadline.

A machined compute core
02 / RENDER

Render

Every node runs the same deterministic kernel, so the same tile seeded the same way produces the same pixels on a 3080 as on a 5090.

Two identical tiles, one carrying a verification grid
03 / VERIFY

Verify

Proof-of-Render re-renders random tiles on a second node and compares the result. A tile that cannot be reproduced is never paid for.

02Network
Live network

Running right now

Every figure below is read from the network API on this page load — not typed into a template.

—Nodes online
—Aggregate TFLOPS
—VRAM online (GB)
—LUMEN minted
—Frames delivered
—Continuous uptime
03Flow
How it works

Three steps from
scene to finished frame

01

Submit

Upload the scene, choose a quality tier, set a spend ceiling. Nothing to install and no node to configure.

02

Split

The scheduler cuts the job into tiles and assigns each one by available VRAM, capability tier and current load across the network.

03

Settle

Verified tiles are paid in LUMEN the moment they land. Anything that fails verification is re-queued and the node loses its stake.

04Output
What comes back

Finished frames

Automotive, product, architecture, character, material and simulation work rendered on the network.

Automotive render
Automotive4K · 512 spp
Product render
Product4K · 256 spp
Architecture render
Architecture6K · 384 spp
Character render
Character4K · 320 spp
Caustics study
Materials2K · 640 spp
Cloth simulation
Simulation4K · cloth
Token

Burn-mint
equilibrium

Render fees burn LUMEN. Verified work mints it. Supply is a function of how much the network actually renders, not of a release schedule.

Burn on one face, mint on the other
Total supply536,870,912
IssuanceBME · per verified tile
SettlementRobinhood Chain
Node collateralStaked LUMEN
Run a node

Your GPU is idle
about 90% of the time

Install the agent Read the spec