Skip to main content
Daydream is an open-source toolkit for world models and real-time AI video built on the Livepeer network. The Daydream gateway routes real-time AI inference jobs - style transfer, image-to-video, StreamDiffusion pipelines - to Livepeer’s decentralized GPU network.
Daydream is a full product with its own documentation. This page covers the gateway aspect relevant to Livepeer developers. For the complete product documentation, visit docs.daydream.live.

​
Start here in 5 minutes

​
What the Daydream Gateway Does

The Daydream gateway is a Livepeer gateway node operated by the Daydream team. It accepts real-time video and AI inference requests and routes them to compatible orchestrators on the Livepeer GPU network. Use the Daydream gateway to:
  • Transform a live camera feed into AI-stylized video in real time using text prompts
  • Run StreamDiffusion pipelines (image-to-image, style transfer) on video streams
  • Experiment with world models and generative AI video workflows
  • Access Livepeer AI inference without running your own gateway node

​
Access Options

The fastest way to experience Daydream is the hosted app at daydream.live. No API key or setup required - open the app, allow camera access, enter a text prompt, and your video is transformed in real time.This is ideal for creators, artists, and anyone who wants to explore real-time AI video without writing code.

Try Daydream

Open the hosted Daydream app - no account required.

​
First Request Example (Portal-Guided API)

Use the Daydream docs as the source of truth for the current API base URL and route, then run a minimal authenticated request: 
DAYDREAM_API_URL="<copy from docs.daydream.live quickstart>"

curl -X POST "$DAYDREAM_API_URL" \
  -H "Authorization: Bearer <DAYDREAM_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "cinematic watercolor skyline at dusk"
  }'
Expected success signal: HTTP 200 response with generated output metadata (for example, an asset URL, stream URL, or job/result payload depending on the selected route).

​
What Daydream Supports

Real-Time Style Transfer

Transform live video with text prompts using StreamDiffusion. Sub-second latency on GPU.

Image-to-Video

Generate video from image inputs using diffusion model pipelines.

World Models

Experiment with generative and interactive world models for video.

Custom Pipelines

BYOC (Bring Your Own Compute) support for custom AI pipelines via the Livepeer gateway.

​
Network Architecture

Daydream runs on Livepeer’s GPU network. When you send a job through the Daydream gateway:
  1. Your request hits the Daydream gateway node
  2. The gateway routes the job to a compatible orchestrator (GPU node)
  3. The orchestrator runs inference and returns the result
  4. The gateway streams the output back to your client
This means inference is distributed, low-latency, and scales with network GPU supply - without Daydream needing to own or provision GPUs directly.

​
Community and Resources

Daydream Product Overview

Full Daydream product context, videos, and ecosystem fit.

Run Your Own Gateway

Build your own gateway node for custom AI video routing.

AI Pipelines Overview

How AI pipelines, ComfyStream, and BYOC work on Livepeer.

Find Gateway Services

Compare all gateway providers.
Last modified on March 9, 2026