Livepeer Launches Subnet for AI Compute to Process Generative Video

(US and Canada) Open video infrastructure network Livepeer has announced the launch of its AI subnet. As the usage of generative AI increases, there's a commensurate need for more adaptable, dynamic access to AI compute capabilities for builders, creators and developers, says Livepeer CEO Doug Petkanics: "AI inference is one of the most costly and compute heavy challenges with generative AI," says Livepeer CEO Doug Petkanics. "Our global network of decentralized nodes can serve the growing demand for AI compute – demand that will only grow exponentially as text-to-video models become embedded across entertainment, social media and gaming."

The Livepeer AI subnet allows hardware providers to earn fees for contributing their GPUs for AI inference jobs such as text-to-image, text-to-video and image-to-video. The subnet is a fork of the main Livepeer network, which enables new AI capabilities while still using the Livepeer protocol for discovery and payments to nodes.

The goal of the subnet launch is to on-board nodes to support Livepeer's new AI capabilities, perform testing and collect data from strategic test partners that benchmark the cost effectiveness of the Livepeer network for performing AI inference at scale, continues Petkanics:

"After validation, the team will merge these capabilities into the core Livepeer clients, add additional job types, and grow the ecosystem around leveraging additional forms of AI-based video compute. The goal is to enable AI developers to bring their own models, weights, fine-tunings, or deploy custom LoRAs on top of existing foundational models within the network. These will give the ultimate flexibility to developers, who can design bespoke workflows to support any type of AI media compute need imaginable."

The team is currently targeting an August launch for the Livepeer AI mainnet, which will allow builders, creators and developers to leverage Livepeer's cost-efficient, reliable and scalable infrastructure for all of their generative AI compute needs.

Livepeer joins the C2PA to fight deepfakes

Livepeer is the first open infrastructure network to join the Coalition for Content Provenance and Authenticity (C2PA), an open technical standard providing publishers, creators, and consumers the ability to trace the origin of different types of media. Other members include TikTok, Adobe, Google, Sony, Intel, BBC, Microsoft and OpenAI.

"When it comes to AI, content provenance is difficult," says Petkanics. "The infrastructure layer is ideally positioned to be a crucial stopgap for misinformation, as it can seamlessly fingerprint content as it moves from creation to editing, distribution and consumption. We're committed to supporting content authenticity wherever possible to rebuild trust and embed verifiable metadata that enables attribution at each stage of the content lifecycle."