Doug Petkanics, is the Co-Founder & CEO of Livepeer a decentralized video streaming network built on the Ethereum blockchain.
Could you share the genesis story behind Livepeer?
I founded Livepeer in 2017 with my long-time collaborator Eric Tang. We’re both entrepreneurial software engineers and had been co-founders before. One of our previous companies was Hyperpublic, an open database and API of local data from around the world. After that company was acquired by Groupon, we then built a company that relied on APIs from the tech giants like Facebook, Google, Pinterest, and Twitter.
The company ended up not working out because none of those platforms remained open, as they were when we were originally building, and our access was abruptly cut off as a result. Our app was also held up for approval by Apple, which hampered our progress. From that point on, we wanted to build outside of walled gardens in an open and decentralized manner. We turned our attention to what was then an emerging concept: blockchain.
We were focused on Ethereum, which we saw as a powerful new paradigm because it enables open-source software where we could use economic incentives to build a community that not only contributes to that software but enjoys the economic upside of its success.
Prior to Livepeer, we built two Ethereum-based protocols: Auction House, a DApp to sell non-fungible assets and TAP, a protocol designed to increase security and transparency in Ethereum transactions.
During this time, we noticed the incredible rise of streaming video, which now accounts for up to 80% of internet traffic. We saw streaming emerge as a tool for creators to communicate directly with their communities. Now we call it the Creator Economy, but at the time we just saw an opportunity to propose an alternative to the closed ecosystems being protected by the Big Tech gatekeepers that owned YouTube and Twitch.
We started Livepeer to rebalance these dynamics and encourage innovations that wouldn’t be possible with traditional, centralized cost structures.
Could you briefly explain what Transcoding is and why it matters for live or on-demand video streaming?
Transcoding is the process of turning a live video feed into streamable packets to distribute to end-users. Transcoding processes video from broadcasters into the various bitrates and formats required by the platforms broadcasters choose.
The process is essential because it’s how raw video is reformatted into the optimal viewing experience, regardless of device or network speed. And it’s only becoming more mission-critical with the arrival of 4K video and ultra HD. In the not-so-distant future, the rise of the metaverse will introduce VR streaming, which will require exponential growth in transcoding.
Why is the Livepeer network more cost-effective to transcode videos than current cloud solutions?
Transcoding is one of the most expensive parts of running a live stream network. It costs about $3 per hour per stream on cloud services. Even if you wanted to own your own media server, it can cost thousands of dollars in licensing fees.
These high costs trigger skyrocketing streaming bills as a platform scales – and can create the sad situation where a startup is so successful that it runs out of money! Or it has to quickly slap on a business model before it has product-market fit, such as selling user data or diluting the user experience with intrusive advertising.
Take PlayDJ.tv as an example. After launching in April 2020, the DJ streaming site’s founders couldn’t afford to keep paying for the infrastructure costs when the site was an instant hit due to the pandemic.
That’s why Amazon turned infrastructure into a business: It’s a very expensive aspect of running online services — especially a livestreaming platform. Monetizing that infrastructure turns an expense into a profit center. And there’s not much competition, which means that there’s no downward pressure on AWS, the biggest transcoder. They can basically set a price and that’s it.
The average cost on the Livepeer.com platform is around 30 cents an hour, so we’re 10x more cost-effective. And the open Livepeer network can be even more cost-effective than that.
We achieve this dramatically different pricing thanks to our network of node operators. We call them orchestrators, and they are usually cryptocurrency miners devoting spare GPU capacity to the network. It’s extremely efficient, as they can still maintain mining operations while also transcoding video, and our technology automatically routes transcoding requests to GPUs with minimal hashrate loss.
And, since we are completely focused on building the decentralized live video infrastructure for streaming applications, developers get reliable and robust infrastructure that ensures the best experience. It’s our core business and that’s much different than paying AWS to transcode your livestreams – and lining the pockets of the same company that owns your biggest competitor, Twitch.
How can users contribute computing power and how are they rewarded?
Orchestrators provide computing power in exchange for fees, which are paid by broadcasters. Orchestrators also receive rewards in our native token, LPT. The formula is based on total minutes of video transcoded, as well as the reliability and quality of individual node operators.
The community can also earn a portion of these incentives as delegators by staking tokens to choose well-performing and reliable orchestrators. We have a great visual primer on that here.
In addition to earning fees, Livepeer mints new token over time, much like Bitcoin and Ethereum block rewards, which are split amongst Delegators and orchestrators in proportion to their total stake relative to others in the network. This grows network ownership amongst those who participate and shrinks it with non-participants.
It also gives Livepeer a large cost advantage over centralized providers, since the value of the token offsets what orchestrators need to charge broadcasters to break even. With traditional centralized video providers, they have to charge you their cost of service for transcoding and distributing video plus a margin.
What’s the process for a developer who wishes to use the Livepeer infrastructure services to transcode and stream videos?
We built our Gateway API to be easy-to-use and straightforward to implement. We encourage any developers interested in exploring Livepeer to check out our documentation or reach out to our dedicated success team to help you get started. For deeper contributions, our entire project is open source and our documentation is editable on GitHub.
We also know all too well how precious every dollar is in the early stages of a startup. So developers can also stream up to 1000 minutes for free each month on Livepeer.
Could you discuss some of the machine learning technologies that are used at Livepeer for both object recognition and song recognition?
In addition to our easy-to-use developer API and affordable transcoding, we’re also working on AI-based features to support content moderation and interactivity.
The one that is in pilot right now with real users is called scene classification, which allows our system to create automatic alerts when the content being streamed may be adult content, violent content, or copyrighted content like Premier League Soccer. This allows the sites to take action either automatically, or to flag the abusive stream to a moderator for review.
Without this technology, it’s incredibly expensive to employ an enormous team of people to moderate every stream. Livepeer can do this work for you while transcoding your video.
What are some of the challenges behind scaling both the transcoding and streaming of videos?
One of Livepeer’s advantages is that it can support transcoding at a significant scale, without the need to pre-provision capacity or have reserved machines standing by to process your streams. The ultimate challenge, however, for any live stream, is that this work needs to be done very quickly in order to ensure that there is no buffering for users around the world.
This is largely a technical execution – can the Livepeer software be fast enough to make sure it can find a transcoding node for your stream, get it transcoded, and failover to another if that first node doesn’t do the job fast enough? There are a number of algorithms and techniques to maximize these chances. And, while the network is performing up to industry-grade transcoding reliability, it will be a constant challenge to improve upon as our network grows.
What’s your vision for Livepeer 5 years from now?
I’m really excited to see the innovations enabled by Livepeer’s open video infrastructure. As the next generation of creator economy services rely less on centralized, expensive infrastructure, I hope to see a thriving ecosystem that’s less beholden to just a few tech companies. As Web3 scales, it would be a shame to simply replicate the existing gatekeeper model.
In five years, I hope to see many truly decentralized options flourish. Ultimately, the world’s open video infrastructure should fulfill the promises of being infinitely scalable, optimally cost-effective, and maximally reliable. If it fulfills these, then the Livepeer software and protocol should work their way into every video workflow, across the majority of devices on the planet over time.
Is there anything else that you would like to share about Livepeer?
Livepeer is both an open community project that anyone can get involved in, as well as a company that works to build video software and services that drive adoption to the Livepeer network. Visit livepeer.org to get involved. We’d love to have you!