Livepeer

**Wireless infrastructure is being designed from the ground up for real-time AI video processing.** Key developments: - 6G networks targeting sub-100ms latency specifically for AI video workloads - Production studios moving AI video processing in-house as competitive infrastructure - Industry consensus: 2026 marks the shift from batch processing to real-time video operations **The challenge:** Real-time AI video requires 1000x more compute than text or images, yet most GPU infrastructure wasn't built for sustained, low-latency video workloads. **What matters now:** Having compute infrastructure that can handle continuous video processing without spiraling costs at scale.

Livepeer has launched real-time AI video processing capabilities, offering significant cost savings of 50-80% compared to traditional cloud services. **Key Features:** - Real-time video-to-video AI processing now available - Built on decentralized infrastructure - Open source and community-powered **Try It Now:** Users can test the technology through [Daydream Live](https://daydream.live), an application built on Livepeer's infrastructure that demonstrates real-time video AI capabilities. The launch represents a practical application of decentralized video infrastructure, making AI video processing more accessible through lower costs while maintaining real-time performance.

The **Lisar team** has completed key development milestones ahead of their public launch: - ✅ **Closed beta testing** - Full lifecycle review completed - ✅ **Live Transparency Dashboard** - Real-time community visibility now active The team is now preparing for **public launch mode**, marking a significant step forward for the project. This progress demonstrates steady development momentum as Lisar moves closer to broader availability.

Livepeer has released new **LISAR SPE release notes** following the successful LiveInfra SPE proposal that passed in July. The update enhances the **Community Arbitrum Node**, providing developers with: - Free, reliable RPC access - Improved uptime and redundancy - Better overall performance This development strengthens Livepeer's commitment to **public goods** and improving the developer experience for builders in the ecosystem. [Read the full release notes](https://forum.livepeer.org/t/lisar-spe-release-notes/3159)

**Livepeer achieves record network fees** as AI processing drives 72% of revenue, marking a 3x year-over-year increase in Q3. **The transformation is complete** - from streaming infrastructure to real-time AI video processing powerhouse. **Key metrics:** - Q3 network fees up 3x YoY - 72% of fees now from AI processing - Projects like DaydreamLiveAI and Agent SPE avatars live **Three competitive advantages:** - Lowest latency for real-time AI - Cost-effective pay-as-you-go scaling - Open, decentralized community ownership **Next phase focuses on:** - Deep productization - World-class latency benchmarking - GPU marketplace scaling - Direct builds for gaming, entertainment, robotics Livepeer positions itself as the infrastructure layer for next-generation real-time AI video applications. [Read the full vision update](https://blog.livepeer.org/a-real-time-update-to-the-livepeer-network-vision/)

**Livepeer's Cloud Team** has rolled out major infrastructure improvements to strengthen the decentralized video network's performance monitoring capabilities. **Key Technical Upgrades:** - Live Stream Input Support (RTMP/FFmpeg) - New Scoring Logic for Live Video Jobs - Integration with MediaMTX and Multi-URI Management These enhancements address a critical gap in **live video pipeline support**, with live video job scores now visible in the Livepeer Explorer. The updates aim to maintain Livepeer's position as the leading reliable and transparent decentralized video compute network. [Read full technical details](https://forum.livepeer.org/t/update-expanding-live-video-metrics-and-testing-capabilities-across-the-livepeer-network/3114)

**Daydream**, a new open-source platform for real-time video-to-video AI transformations, has launched on Livepeer's decentralized infrastructure. The platform enables real-time AI video processing and is now live at [daydream.live](https://daydream.live). Key features: - Community-powered development - Open source architecture - Built on Livepeer's decentralized network - Real-time video AI capabilities Livepeer is supporting this initiative through delegator workshops and network infrastructure support. *Try the early preview to explore real-time AI video transformations.*

Peter from AI SPE discusses the evolution of real-time AI video on Livepeer in the latest Fireside episode. Key developments include: - Transition to Bring Your Own Container (BYOC) architecture - Introduction of ComfyStream for customizable AI pipelines - New community-driven development approach This follows recent episodes featuring UFO's decentralized radio integration and earlier discussions with Streamplace, ORIGIN STRIES, and Eliza agents about AI-powered video innovations. [Watch the full episode](https://youtu.be/wFSs8kfvsYA)

Livepeer cofounder Doug Petkanics outlines the network's expansion into real-time AI video processing. The initiative promises to enable live video manipulation as simple as typing ChatGPT prompts. Key developments: - Network aims to provide ultra-low latency, heavy compute, and global reach - New features include AI VTubers, interactive livestreams, and real-time translation - Daydream platform launches as proof-of-concept - Treasury funding supports development of new applications The network expects significant increase in usage fees for node operators and delegators through AI video processing capabilities. [Try Daydream](https://daydream.live/) [Read Full Article](https://blog.livepeer.org/daydream-live-a-glimpse-into-the-future-of-realtime-ai-video-on-livepeer/)

Daydream is set to demonstrate ComfyStream at ComfyCon, the first official global ComfyUI conference in Shanghai on March 29-30, 2024. The presentation will showcase how ComfyStream integrates with the Livepeer ecosystem to enable real-time AI video workflows. This follows their recent collaboration at NYU, where students explored Real-Time Video AI applications with Livepeer and ComfyUI. Join the event: [ComfyCon Registration](https://lu.ma/ComfyCon)