Prerequisites
- Dependencies (binary):
curl,coreutils,gnupg2(Linux) orgnupg(macOS). - Docker (Docker install): Install Docker and, for GPU support, NVIDIA Container Toolkit.
- GPU (transcoding): NVIDIA driver; for Linux GPU binary or build from source, CUDA Toolkit (e.g. v12).
Binary release
Replace<RELEASE_VERSION> with a release tag (e.g. v0.5.35) and <ARCH> with amd64 or arm64.
Darwin (macOS)
Linux
Linux (GPU / NVIDIA)
Use the GPU build if you will run transcoding on NVIDIA GPUs. Requires CUDA 12.Windows
Livepeer does not provide automatic updates. You can update manually or use a community script (e.g. Forum: Bash script to update Livepeer).
Docker
Images are published to Docker Hub. For the latest pre-release:--gpus all:
To use the CLI from the host:
Build from source
- System deps (Ubuntu 20.04+):
- Go: Install Go.
- Clone and build:
For GPU support on Linux, install the CUDA Toolkit and set
LIBRARY_PATH; install clang if required by the ffmpeg script.Activate
After installation, run the node in orchestrator mode, fund your account with ETH and LPT, then activate:- Run livepeer_cli and choose the multi-step “become an orchestrator” option.
- Set block reward cut and fee share (percent kept vs shared with delegators).
- Set pixels per unit and price per unit (wei).
- Set service address (must match
-serviceAddrused when starting the node). - Bond LPT (amount in LPTU, e.g. 5000000000000000000 for 5 LPT).
- Wait for transactions to confirm; your orchestrator joins the active set in the next round.