Choose the quickstart for your build goal. If you are unsure which applies, use the navigator to compare paths first.
AI inference
Goal: Make a working API call to a Livepeer AI pipeline. Time: 10 minutes. Prerequisites: API key from Livepeer Studio. No GPU. No local node. What you build: APOST /text-to-image call that returns a generated image, using curl or an SDK of your choice.
SDK options: livepeer (npm v3.5.0 / PyPI), @livepeer/ai (alpha, AI-only).
AI Quickstart
First API call, error handling, and next steps in under 10 minutes.
Real-time AI video
Goal: Apply a ComfyUI workflow to a live webcam or video stream. Time: 30-60 minutes (includes environment setup). Prerequisites:- NVIDIA GPU (RTX 3090 or better recommended)
- ComfyUI installed and working
- Python 3.12, CUDA 12.5+, Miniconda
ComfyStream Quickstart
ComfyStream install, SD1.5 workflow, and webcam-to-processed-stream in 30 minutes.
Video transcoding and streaming
Goal: Create a livestream or transcode a video asset using the Studio API. Time: 15 minutes. Prerequisites: API key from Livepeer Studio. SDK installed. What you build: A stream with an RTMP ingest URL and an HLS playback URL, tested end-to-end. SDK install:Transcoding Quickstart
Create a stream, get the ingest URL, test with OBS or ffmpeg.
OSS contribution
Goal: Get the Livepeer codebase running locally and make your first contribution. Time: 1-2 hours (includes testnet setup). Prerequisites: Go 1.21+ for go-livepeer; Python 3.12 for ai-runner or ComfyStream; Docker. What you build: A local testnet with one orchestrator and one gateway running, capable of processing test jobs.Contributor Quickstart
Clone go-livepeer, start a local testnet, and submit your first test job.
Not sure which path fits?
Related pages
Navigator
Full path comparison with decision framework and per-path descriptions.
Workload Fit
Studio API vs self-hosted gateway vs BYOC vs ComfyStream — when to use each.