Daydream.live: A Glimpse into the Future of Realtime AI Video on Livepeer

The Livepeer project has always been driven by a singular vision: to build the world's open video infrastructure. As technology evolves, so do the possibilities—and one of the most exciting frontiers we've entered is the application of AI to live video. Imagine personalizing a live video stream as effortlessly as typing a prompt into ChatGPT—that’s the fun and easy experience you can find in Daydream. This playful new app instantly turns any webcam feed into an AI-generated canvas.

This is not just a natural progression. It’s a massive opportunity. Realtime AI video workflows demand intense compute resources, ultra-low latency, and a sophisticated global video infrastructure—core strengths of the Livepeer network, enabled by its token, LPT. Through our Cascade initiative, we’ve positioned ourselves to meet this moment head-on, aiming to become the infrastructure layer for the future of realtime video.

But as with every technological shift, adoption begins with curiosity. It begins with play.

Introducing Daydream.live

Daydream.live is a realtime AI video playground that lets users type a prompt and instantly see their live video stream transformed in response. What once felt like a static, limited interface—choosing between a few preset filters—is now a generative, infinite canvas for self-expression.

Using just a webcam or smartphone, anyone can experience the magic of realtime video AI. It’s fast. It’s fun. It’s inspiring. And it’s built to spark that a-ha moment that comes when you realize: “This is what the future of video feels like.”

A Dual-Pronged Approach to Seeding the Market

Livepeer is approaching this emerging market from two complementary directions:

1. Empowering Developers: We’re working closely with the ComfyUI community—the leading open source AI project with a strong community of millions around the world—to unlock new creative workflows using Comfystream, our integration that brings live video into ComfyUI pipelines. This community is helping invent the building blocks of tomorrow’s real-time video experiences.
2. Inspiring Users: Daydream.live is designed for creators, hobbyists, and technophiles—those who want to feel the power of live AI. Whether they’re generating clips for use in other tools like Sora, streaming directly to Twitch, or just exploring new forms of self-expression, Daydream is a fun and accessible entry point into the world of realtime AI.

What We’re Learning from the Early Community

Even in its early stages, the Daydream community is surfacing powerful signals:

• Rapid Creative Prototyping: Users report that this is the fastest way to prototype creative video concepts. Instead of iterating through prompt-image-review cycles, they can simply go live and explore ideas in motion. It's not just a time saver, but it’s also enabling video content that wasn’t possible before. People have been uploading Daydream videos into Sora or Runway, creating stunning results that cannot be achieved otherwise
• Avatars: Daydream's real time video transformation is the easiest way to embody an avatar as your online persona. It's like the magic mirror that turns you into anything you want to be.

These are just two examples of early paths that could lead to product-market fit. The key insight is that when you give thousands of people a new tool that feels like a toy, they begin to reveal the future use cases. And once that happens, the ecosystem has a clear direction.

The Flywheel of Innovation and Network Usage

The realtime AI video opportunity aims to dramatically increase usage of the Livepeer network, and hence fees flowing to node operators and delegators, an order of magnitude beyond what streaming and transcoding alone have provided.

Here’s how it fits together:

• Builders get inspired by what users want and prototype new workflows.
• The community treasury can fund independent teams to turn these prototypes into polished applications.
• Each new use case drives compute usage, and thus fees, on the Livepeer network.
• The network optimizes for performance, low latency, and cost-effectiveness across an expanding set of live AI workflows.

Daydream is just one workflow, built using Dreamshaper as the base model. While appearing simple on the surface, it is actually a complex workflow including depth mapping, image generation, and face tracking - the component parts are as interesting as the productization. This is a demonstration of how you can use multiple realtime capabilities to craft an effective product.But it’s also just the beginning. New workflows are coming—from live translation and dubbing, to VTubers, interactive livestreams, AI agents, closed captioning, security monitoring, and more. Each new addition is composable and additive, creating a growing network effect.

Unlike generic GPU infrastructure, Livepeer offers a tailored live video stack, performant workflow primitives, and a built-in community of users ready to try the next big thing. That’s what makes this infrastructure valuable.

Daydream as a Signal and Showcase

Daydream today is not the product. It’s the signal.

It showcases what’s possible when AI meets live video. It inspires developers and gives the Livepeer community valuable insight into what the world wants to do with this new capability. It sets the stage for dozens of future applications that embed realtime AI video functionality in more focused contexts.

And most importantly, it brings us closer to Livepeer’s long-standing mission: to be the infrastructure layer for the video applications of tomorrow.

Because in the future, AI will run on video—just like it’s already running on text and images. And when it does, the Livepeer network, node operators, and community will be ready. Get a glimpse at that Future.