Skip to main content
A pool lets GPU compute contribute to the Livepeer network without running a full Orchestrator. The pool operator handles staking, on-chain registration, reward calling, and job routing. Workers provide GPU hardware and earn for the work it processes. This is the lowest-barrier entry point for GPU owners who want to monetise compute without taking on protocol operations.
Pools are community-run and each pool operates independently. Vet any pool carefully before connecting hardware.

​
Pool Architecture

An Orchestrator pool is a single registered Orchestrator node that aggregates compute capacity from multiple GPU contributors. From the network’s perspective, it appears as one Orchestrator with one stake. From a worker’s perspective, the GPU is one contributor behind that Orchestrator. The pool operator handles:
  • Holding and managing the stake
  • Setting prices and advertising capabilities
  • Receiving work from Gateways
  • Routing segments to worker GPUs for processing
  • Distributing payouts to pool members
A worker:
  • Runs a transcoding worker process (go-livepeer in transcoder mode, or the pool’s own client)
  • Processes the segments routed to the GPU
  • Receives off-chain payouts based on completed work
The pool operator tracks individual worker contributions in pool-side systems. Payouts are made directly by the pool operator, usually in ETH and sometimes in pool-specific tokens.

​
Pool Worker vs Solo Orchestrator

​
Step 1: Choose a Pool

The Livepeer community currently has a small number of active public pools. Titan Node operates the most visible one, with a public dashboard and documented setup process.

Titan Node Pool

Community-run video transcoding pool. Docker-based worker client, a public dashboard at app.titan-node.com, and payout tracking managed by the pool operator.

Livepeer Discord #orchestrators

The active community channel for finding pools, asking about current payout terms, and connecting with other operators.

​
Evaluating a pool before joining

Pool terms vary. Before connecting hardware, confirm:
Reputable pools publish their payout formula. Some pay per segment transcoded; others pay proportionally to GPU time contributed. Ask for the formula and compare it to the expected return from a solo node.Common payout models:
  • Per-segment - a fixed amount per segment processed
  • Proportional share - the worker’s earnings match their share of total pool work
  • Pool token - some pools issue a native token that is exchangeable for ETH or worth holding
Always confirm what is actually paid before committing hardware.
Most pools pay weekly or monthly, with a minimum balance before payout triggers. Ask the pool operator what the threshold is and how long a new contributor typically takes to reach it.A high minimum threshold combined with low network demand extends the wait for a first payout. Factor that into the decision.
Legitimate pools provide clear technical requirements: GPU type, minimum VRAM, operating system, and network bandwidth. A legitimate pool never asks for a keystore file or private key.A request for a keystore or private key is a scam. Stop there.
Check when the pool’s GitHub repository or documentation was last updated. Check Discord presence and community activity. A pool with stale tooling often has weak Gateway demand.

​
Step 2: Connect Your GPU

Pool operators each have their own worker client setup. The general connection model is consistent across pools. Most pools provide a Docker image. This is the simplest path and keeps setup focused on worker configuration.
Docker worker setup
# Clone the pool's repository
git clone https://github.com/Titan-Node/Titan-Node-Pool.git
cd Titan-Node-Pool/docker

# Review and edit docker-compose.yml with your settings
docker compose up -d
The container connects to the pool operator’s Orchestrator, receives video segments, processes them with the local GPU, and reports back completed work.

Titan Node Docker Setup

Docker Compose configuration and worker setup steps for the Titan pool.

​
Option B: go-livepeer transcoder mode

Run go-livepeer directly in transcoder mode and point it at the pool operator’s Orchestrator address:
go-livepeer worker command
livepeer \
    -transcoder \
    -orchAddr <pool-orchestrator-address>:8935 \
    -nvidia 0 \
    -maxSessions 10
The pool operator provides their Orchestrator address and any required authentication configuration.
Pass -transcoder only when joining a pool. That keeps the machine in worker mode and leaves on-chain registration with the pool operator.

​
Option C: Cloud GPU

Cloud GPU instances (RunPod, Lambda Labs, AWS) also work as pool workers. This path works only when compute cost stays below expected earnings. Check current network demand in Discord before committing to a cloud GPU rental.

​
Step 3: Verify Work Is Arriving

Once connected, transcoding activity should appear in worker logs within a few minutes during active pool routing windows. In go-livepeer transcoder mode, look for:
Worker activity log
Received transcode request  segment=0 ...
Transcoded segment          segment=0 ...
No activity after 10-15 minutes:
  1. Verify the pool Orchestrator endpoint address is correct
  2. Confirm GPU is visible: nvidia-smi should show the GPU and go-livepeer should log a GPU detection line at startup
  3. Check that port 8935 is reachable from the pool Orchestrator (some pools require inbound connectivity from the worker)
  4. Ask in the pool’s Discord channel - quiet periods occur when network demand is low

​
Step 4: Track Your Earnings

Pool payouts are off-chain. Worker earnings are tracked by the pool operator through pool-specific dashboards, bots, or statements. Each pool uses its own tracking mechanism:
  • Titan Node Pool: public dashboard at app.titan-node.com shows per-worker stats and pending payout
  • Most pools: Discord bot or web dashboard showing segment count and pending balance
Keep a record of expected versus actual payouts. Discrepancies should be raised in the pool’s community channel.

​
Frequently Asked Questions

The pool operator handles LPT staking, token ownership, and on-chain actions. Pool workers provide GPU compute only.
On-chain registration stays with the Orchestrator node. Individual worker contributions are tracked within the pool operator’s own systems and any dashboard they provide.
Transcoding fees are paid in ETH. LPT inflation rewards accrue to the staking Orchestrator. Some pools also share a portion of LPT rewards with workers. Ask the pool operator about their current policy.
AI inference via the AI subnet requires capability registration tied to the Orchestrator node. As of early 2026, publicly documented pools primarily handle video transcoding. Running AI inference workloads and earning from them currently requires a solo Orchestrator path. Check Discord for any AI-focused pools that have launched.
Yes. Pass multiple GPU IDs when starting the worker (e.g. -nvidia 0,1,2). The pool Orchestrator sees the worker as a single endpoint with higher capacity. Most pools support this without special configuration.
Livepeer transcoding pools work differently from proof-of-work mining pools. Earnings are tied directly to actual video segments processed, so returns follow network demand instead of a fixed hardware-output curve. The primary variable is transcoding demand from applications routing video through the Livepeer network.

Setup Options

Compare pool worker, solo Orchestrator, and split setup before choosing a path.

Solo Setup Guide

Install go-livepeer, register on-chain, and start earning directly as a solo Orchestrator.

Operating Rationale

Compare pool worker earnings against running a solo Orchestrator with the full cost and revenue breakdown.

Community Pools

Directory of active community-run pools, dashboards, and worker setup resources.
Last modified on March 16, 2026