Skip to main content
After installing go-livepeer, configure the node so gateways can discover you and send work. This page covers core flags and optional config files.

Core configuration

Network and wallet

  • -ethUrl — Arbitrum RPC URL. See Connect to Arbitrum.
  • -ethAcctAddr — (Optional) ETH account address if you already have a keystore.
  • -network — Use arbitrum-one-mainnet for production.

Orchestrator identity and pricing

  • -serviceAddr — Public hostname or IP and port (e.g. orchestrator.example.com:8935). Prefer a hostname you control; changing it later requires an on-chain update.
  • -pricePerUnit — Price in wei per pixel (transcoding). Required at startup; can be updated later via CLI or on-chain.
  • -pixelsPerUnit — Pixels per pricing unit (default 1). Affects how price is calculated per segment.

GPU and mode

  • -orchestrator — Run in orchestrator mode.
  • -transcoder — Enable GPU transcoding.
  • -nvidia — Comma-separated NVIDIA GPU IDs (e.g. 0,1). Omit for CPU-only.
Example minimal run:

Optional: transcoding options

You can define transcoding profiles in a JSON file and reference them. Example structure:
transcodingOptions.json
[ { "name": "240p", "width": 426, "height": 240, "bitrate": 250000, "fps": 30, "profile": "h264constrainedhigh" }, { "name": "720p", "width": 1280, "height": 720, "bitrate": 3000000, "fps": 30, "profile": "h264constrainedhigh" } ]

Optional: AI models (AI orchestrators)

If you run AI inference, create an aiModels.json to specify pipelines and models:
aiModels.json
[ { "pipeline": "text-to-image", "model_id": "stabilityai/sd-turbo", "warm": true } ]
See AI pipelines and Developers BYOC for full AI setup.

Verification

After starting the node:
  • Logs show Listening for RPC on :8935 and no RPC errors.
  • Explorer shows your node after activation.
  • Under load, GPU usage (e.g. nvidia-smi) increases when jobs are processed.

See also

Last modified on February 24, 2026