Compiled by Alex Liu, Foresight News
Launched in 2017, Livepeer is the first fully decentralized live video streaming network protocol. The platform aims to provide a blockchain-based, cost-effective alternative to traditional centralized broadcasting. It aims to revolutionize the fast-growing live video streaming and broadcasting industry by allowing producers to submit their work on the platform, which is then responsible for reformatting and distributing the content to users and streaming platforms, introducing a decentralized ecosystem.
Simply put, through the DePin facility, Livepeer can seamlessly integrate video content into applications in a decentralized manner at a fraction of the cost of traditional solutions.
The popularity of the DePin track surged at the beginning of this year, and LPT also caught up with the growth, with the token price doubling compared to the beginning of the year. As a leek who bought 10 yuan a year ago and sold all at 9.8, the editor decided to learn from the pain and carefully study Livepeer's new move - launching an AI subnet.
Livepeer launches AI subnet to innovate and change
In the era of generative AI, video creation has ushered in new changes.
The field of generative video has grown rapidly since Open AI’s Sora demo showed the possibility of creating videos by inputting text prompts. The open source AI video model Stable Diffusion has surpassed 10 million users in just two months. However, the landscape of generative AI video tools faces daunting challenges. The $49 billion GPU market is controlled by a handful of global internet monopolies, such as NVIDIA, Microsoft Azure, and Amazon Web Services (AWS), driving up prices and creating a global AI computing bottleneck.
Therefore, Livepeer launched the Livepeer AI Subnet: the first decentralized video processing network with AI computing power. The Livepeer AI Subnet solves the structural problems of centralized AI computing by leveraging Livepeer's open network of thousands of GPUs to provide low-cost, high-performance processing services. Based on Livepeer's decentralized video processing network architecture, the subnet provides globally accessible, affordable open video infrastructure and is infinitely scalable through blockchain token economic incentives.
What exactly is the Livepeer AI subnet?
The AI subnet is a branch of the Livepeer video infrastructure network that provides a sandbox environment for secure development and testing of new decentralized AI media processing markets and tools. While the Livepeer network will continue to focus on video transcoding and computing, the Livepeer AI subnet will meet the growing demand for AI computing, handle tasks such as upscaling, subtitle generation and recognition, and support developers to run models for specific video and media tasks.
The subnet allows video developers to add a range of generative AI capabilities to their applications, such as text-to-image, image-to-image, and image-to-video conversion.
This AI generated output is from Tsunameme.ai - the first demo built on the Livepeer AI subnet. It uses the text-to-image and image-to-video pipelines. Try generating your own AI media using the Livepeer beta at https://tsunameme.ai
Reasons for establishing Livepeer AI subnet
AI video tools have lowered the threshold for creation, allowing anyone to create footage that would otherwise require a venue, a professional team, and hours of editing with just a few text commands. As these tools become more popular, the global centralized AI computing bottleneck will be further exacerbated. In addition, decentralized AI infrastructure can also solve the single point failure risks inherent in highly centralized server networks and the trust and authenticity crisis caused by AI-generated content.
The Livepeer AI Subnet provides options for creating a sustainable and profitable open AI video infrastructure by offering globally accessible ultra-low-cost infrastructure, an open and permissionless AI media marketplace, and content verification and authenticity solutions.
How Livepeer AI subnet works
Livepeer uses a decentralized pay-per-task model that allows developers to submit and pay for tasks on demand without having to reserve expensive computing capacity. Developers can set the price they are willing to pay based on the required performance and network availability.
The two key components of the Livepeer AI network architecture are:
AI Coordinator Nodes: These nodes execute AI tasks, keep AI models “warm” on their GPUs for immediate processing, and can dynamically load models to optimize response time and resource utilization.
AI Gateway Nodes: These nodes manage the task flow and assign tasks to appropriate coordination nodes based on capabilities and current load, ensuring efficient task distribution and system scalability.
This diagram illustrates how Livepeer distributes tasks to a distributed network of GPUs based on efficiency, rather than directing AI processing requests through a centralized server.
Unlimited Scalability
The Livepeer AI network infrastructure is designed to be infinitely scalable, allowing for easy integration of additional coordination and gateway nodes as needed. AI models are executed via a dedicated AI-runner Docker image, simplifying deployment and enhancing scalability for new pipelines. Future development will further improve performance and expand the capabilities of the container to support increasingly complex AI models and custom user-defined pipelines.
Technical workflow for processing tasks on the AI subnet. The gateway node passes the task to the orchestrator, which may be running multiple AI-Runner Docker containers of the same or different pipelines. These pipelines may already have the requested models, or can load them dynamically as needed.
Participate in the Livepeer AI subnet
Hardware providers: earn fees by contributing GPUs
Existing Livepeer coordinators can set up and run AI coordinator nodes to perform text-to-image, image-to-image, and image-to-video inference tasks, increasing their existing transcoding revenue.
Developers: Introduce models into the network as AI-Workers
Developers can define and deploy custom pipelines and workflows to ensure their applications are at the forefront of AI and video technology. Developers can also set up AI gateway nodes to test and improve their applications and access APIs for AI tasks.
The launch of the Livepeer AI Subnet marks a significant milestone for the project and the next step in Livepeer’s mission to provide a global, open video infrastructure. With generative AI set to dramatically increase the amount of video content created in the coming years, the Livepeer network aims to ensure it has the capacity to support this wave of growth.
