Text-to-video generative AI makes video creation accessible: Livepeer CEO

In an exclusive interview with Invezz, Doug Petkanics, CEO of Livepeer, shares insights on AI and generative video.

How does Livepeer’s GPU-on-demand network facilitate cost-effective rendering of generative video through text-to-video AI?

Compared to traditional cloud GPU rendering options, Livepeer’s key advantages are its decentralized pay-as-you-go model providing elastic GPU access, open and transparent marketplace pricing mechanics, and specialization for efficient video workload processing.

This positions it well for cost-effectively rendering the latest generative video AI at scale.

Text-to-video AI models, compared to traditional large language models and generative image diffusion models, are extremely computationally intensive and require significant GPU power.

Livepeer provides access to a distributed network of GPU miners who contribute their GPU resources in exchange for earning the LPT token. This allows Livepeer to tap into a large pool of GPU cycles on-demand without having to own and maintain expensive GPU hardware.

And because users only pay for the GPU resources they use through the Livepeer marketplace, they can scale compute needs based on demand without upfront hardware investments or idle capacity costs.

Token-based bootstrapping incentives for GPU providers also help offset the costs that they need to charge to end users, meaning it can be more cost effective than general cloud GPU providers for generative video.

What potential do you see for text-to-video generative AI in revolutionizing industries, and how does Livepeer plan to capitalize on this?

Text-to-video generative AI technology dramatically lowers the amount of resources needed for video creation, which makes it accessible to far more people.

Video content that may have previously required considerable amounts of time and money to produce and edit can now be generated for a fraction of the original cost and at a much more rapid pace.

This has significant implications for any field where video plays a significant role – but at this stage, this technology will likely have the most impact on creative and pre-production processes in the entertainment, advertising, and marketing industries.

People in these spaces can use text-to-video generative AI to quickly and cost-effectively visualize and share ideas – for example, storyboarding for TV shows and films, or creating visual drafts of advertisements.

As this technology continues to develop, work will increasingly be accomplished by infrastructure networks like Livepeer that offer the compute supply needed to bridge the gap between creative concepts and video outputs.

How does Livepeer envision leveraging blockchain technology to enhance the development and deployment of AI-driven video applications?

There are many elements to AI development, but the three that typically require substantial compute power are Training, Fine Tuning, and Inference.

Inference – the act of taking an already trained and tuned model, and having it produce outputs or make predictions based on different sets of inputs – is where distributed networks, including those coordinated by blockchains, can be a particularly effective solution.

Each node operator can choose to load given models onto their GPUs and compete to perform inference tasks. Just like in the Livepeer transcoding network, users can submit these tasks to Livepeer and receive the benefits of open market competitive pricing that can leverage currently idle GPU power.

This is potentially game-changing since inference jobs are often performed over and over again millions of times.

As CEO, how do you navigate the challenges and opportunities presented by the rapidly evolving landscape of AI and crypto?

I’m very fortunate to be working and spending time at the cutting edge of these transformative technologies. I firmly believe that they have the potential to impact society faster and in greater ways than other innovations that have come before them.

Since we founded Livepeer seven years ago, we have always navigated by prioritizing innovation and technical discovery alongside transparent building and communication.

We also recognize the importance of collaborating with others operating in this space. Maintaining an open and curious approach as part of this innovative landscape has served us well along this journey.

We aim to continuously leverage and contribute to the latest technologies to see how we can be a part of delivering major impacts to society based on what we observe in the world and in the market.

What’s Livepeer’s long-term vision for integrating AI into its platform, and how do you anticipate this shaping the future of online video content?

In line with Livepeer’s trajectory over the past 7 years, we aim to demonstrate tangible, practical, open-source software and network capabilities within the AI realm.

We plan to accelerate progress by forking/branching the node software to incorporate new capabilities into both the orchestrator (supply side) node and broadcaster (demand side) node.

Livepeer’s open media server, Catalyst, should facilitate an interface for requesting and consuming these generative video tasks.

As a first step, we’ve identified a specific initial use case for an additional job type beyond video transcoding: AI-based Generative Video, supported by AI upscaling and frame interpolation.

Promising open models like Stable Video Diffusion are rapidly evolving in this domain.

We will eventually seek to harness Livepeer’s cost-effective open compute network to offer services to other applications and prove Livepeer’s cost-effectiveness.

After validating this approach, we will expand the ecosystem around leveraging additional forms of AI-based video computing.

How will this shape the future of online video content?

Text-to-video generative AI will eventually lead to an abundance of content – anyone and everyone will be able to create content with prompts.

This will make effective curation of this content more important than ever. There will simply be so much content out there that distinguishing signal from noise will only get more and more valuable.

There is a massive opportunity to use blockchain technology to crowdsource this curation process.

One of the superpowers of blockchain is that it provides an infrastructure for fast global coordination.

So it’s likely that we’ll see on-chain communities that exist to curate the good and valuable outputs of content generation – for example, which music video will be put to a song, or which video clips get pushed to a larger audience.

Blockchain also enables reward distribution to those who participate in this curation process.

The post Text-to-video generative AI makes video creation accessible: Livepeer CEO appeared first on Invezz