Start with the operating model that fits your hardware, stake, and tolerance for infrastructure work. A pool is the shortest path. A solo node gives you control and the full operating burden.
Do not buy tokens or new hardware before deciding which role you actually want to run.
Fastest path
Connect as a worker behind an existing Orchestrator. The operator handles registration, LPT, routing, and payouts; you provide GPU compute and receive off-chain earnings under the pool’s terms.
Best with 24 GB+ VRAM
Serve inference workloads where capability, price, latency, and uptime matter more than active-set stake.
Full operator path
Operate go-livepeer, publish your service address, manage the wallet, set prices, and monitor rewards.
These are the practical requirements that apply before protocol configuration.
NVIDIA is the supported hardware-accelerated path. Confirm the host can see the card with nvidia-smi.
Use Linux for production GPU workloads. AI also needs Docker, CUDA 12+, and NVIDIA Container Toolkit.
Use stable, low-latency internet. Solo nodes need a public domain or static IP and an open service port.
Account for electricity, storage, bandwidth, and maintenance. Work and earnings are not guaranteed.
Solo on-chain nodes need ETH on Arbitrum One for activation, reward calls, ticket redemption, and ongoing gas.
Official Arbitrum BridgeUse Arbitrum One, not Ethereum mainnet, for the operator wallet’s gas. Confirm the network and destination address before bridging or withdrawing funds, keep an ETH buffer for ongoing transactions, and never paste a private key into a website.
Solo video operators need enough self-stake and delegated LPT to enter the active orchestrator set. Pool workers do not manage stake.
View active orchestrators