UploadCare CEO Igor Debatur on How Poor Video File Management Hurts Website Performance

Hey guys, thanks for the throw. We're here with Igor Dior, who's CEO for upload care and we're talking about the impact that video is having on website performance. 'cause I think everybody in this brother these days has got video on their websites, but I'm not sure we all understand the implications and especially when it comes to performance.

Igor, welcome to the show. Hi there. Hi Mike.

What is going on here? It does seem to me every site now has some video component to it, but do we really understand how to optimize the performance of those sites? 'cause there's attacks that comes with the video and it seems to be in the form of performance, but I'm not quite clear that we even know how to measure that much less manage it.

Uh, it's the problem should, should be simple already in 2025, but it's not as simple as we want it to be. So everything regarding videos, it starts when people want to upload something and usually videos are recorded as more than devices. They wait a ton.

So there are huge video files not compressed and then you need to upload them somewhere. So it's the first potential problem when they're uploaded. The people who build a website or platform, they want to compress them and ideally to deliver on a proper format, depending on the browser, depending on the web internet connection or depending on the screen size.

If you want to deliver them to an iPhone, it's one resolution. If you want to deliver them at large display, it's a different game. And also you don't want to deliver them at once.

You want to deliver them chunk by chunk. You want to stream videos and if you want to build all of that, it's still quite tricky nowadays. And unless you build everything, every part, the videos takes long to download and uh, people spend a lot of time to just to view them.

And yet video is core now to the website experience, even for the average consumer somewhere, not just professionals. And, um, do we really understand what we're doing here? Because as far as I can tell, everything from your Google page rank now to, um, how many seconds people will give you before they leave your website all comes down to performance.

That's right. Unless you optimize your videos and images and all kinds of media content, you won't win the race. You will be less, your ranking won't be as good just because the website doesn't perform as well.

Also, when we, people will still visit your website, they will definitely leave if they fail, uh, to view the content you want to show to them. So how are people managing this today and what should they be doing differently? They're usually using some platforms.

In some cases it'll be a combination of some open source software that you use to get these videos from users. Could be an open source file uploader, then you put them to some AWS storage, you put some Lambda functions or you put some platform to encode these videos. And then you find and uh, integrate some open source video player just to render them and browser for people that to view them.

It's one way. Another way is to generate some code with lms, but basically when you generate all the code, you are doing more or less the same. You find some open source code that was rewritten for your use case and you have to deploy and maintain all of this.

Another way is just to use a redate platform. So there's several platforms available that help you. They encode videos, they're integrated with player and they try to deliver everything as a, in a proper format optimized, uh, for your connection, for your display.

Uh, I'm CEO of one of these companies. So how many folks are kinda, for lack of a better phrase, trying to roll their own here versus relying on a platform? I would say that we, we started uh, about a decade ago and when we started, I would think that maybe 30% of people are looking for some radio made solutions.

Other tried to build stuff by their own. Now this has dramatically changed. People will use their time, they want something that just works.

They don't want to spend time building this from scratch or integrate some open source libraries because when you use open source or when you generate some code by LLMs, you still have to maintain this code. It takes time. It's uh, usually a burden for developers.

So I like, uh, the joke, but it's not a joke. It's a fact that our best customers are CTOs who tried to build some file management video and code and media and coding platforms. They do not want to do this again.

Never. Mm-hmm. Um, as you kinda think through this whole thing a little bit, um, how do I, is storage part of this issue?

I mean, am I putting all this stuff S3 buckets and trying to call it from the cloud? Am I putting it locally? Am I caching on various websites?

Is this part of the process and is that something the platform entirely handles for me? Or how much do I have to think about this? Uh, when you look at the iceberg from the top, it looks very simple.

You just place everything in S3 and this is it. It's infinitely scalable. Price is decent.

You can store any amount of videos you want, but the problem starts where you need to generate, uh, versions, derivatives of these files for different devices, different resolutions, uh, encoded with different uh, codex and then it just multiplies. One way is to just generate versions for every screen, every size and to store all of this. Another version, uh, how our platform, how upload care is built.

We store the source file and we generate versions on the fly. So if the video video is not large and we have different viewers with different demands, we generate versions for all of them. And you do not need to think about these versions, we just generate them on the fly for via requests by users.

And how does that avoid the performance tax we were talking about? 'cause it sounds like maybe I'm just shifting it to your platform or do you have some way of optimizing the, the management of those video files On my behalf, We manage them on our customer's behalf and we generate versions that are the perfect fit for the specific user who wants to view this video on the fly. Um, you wanna explain a little bit about how that works?

What's the magic behind that? The magic is that we have a set of algorithms that are quite fast. It's not just a basic f of impact configured, it's something, uh, more advanced and scalable.

Then we generate, uh, the version based on the user request and what we know about this user, uh, screen size, uh, connection, speed and other parameters. And we generate this version and we deliver optimized video instead of delivering the source content. Alright.

And when you say algorithms, is that AI or a different form of algorithms? Exactly. What are we talking about?

Video and coding. It's not ai, it's just a set of like normal, uh, algorithms because you don't want to use AI for everything. And also it's not that fast.

So when you want to do deliver videos at scale, usually it's not ai, it's just a set of algorithms. If you want to apply some more advanced, uh, transformations for videos, to change backgrounds or to do something else, you want to use ai. But it's not in real time.

It's uh, it works with rest API. So you send the request, then you wait for a result for images, it's different though. So if you want to optimize images, even based on ai, it can be done in real time videos.

They're just more resource, uh, not not effective resource demanding. Are we in the age of AI creating more videos than ever? Have we gotten to that point?

I know we create a lot of images, but how soon will most of the videos we're watching be generated using ai? Plenty of videos are generated using AI and in advertisement. I think, I don't know the percentage, but I would say that maybe 30 persons or more are of advertisement videos are generated with AI or at least modified with AI because it's just quicker, simpler.

And in many cases people do not see the difference or if they see the difference they do not care. Hmm. So one point will websites start to keel over from the load because we're gonna create more video than ever.

And if I don't have some sort of service to manage that and I'm trying to do it on my own, isn't it just a question of time before I run into all the issues we just described. You'll definitely run, uh, to all the issues. And when people think about building some file management, uh, file management and delivery platform, they think that it consists of about five, six parts and it's uh, tens of parts.

It's 50, 60, maybe even one 100 parts because you built it, you built a file uploading part storage, part delivery part and the player sounds simple. But then what about, uh, scalability, resilience of the platform security and changing check check files for NSFW content? What if someone uploads some really dark stuff to the platform and start sharing it to others?

People will have to think about all of this and that's why I wouldn't generate video management platform with l LMS so far. And I wouldn't use some open source software just because it's too tricky. Videos wait too much and it's so much easier just to use some platform.

If I wasn't a CO of hub upload care and I built something else, I would've definitely used the platform 100%. So you mentioned that there are other services besides yourselves, but one differentiate one service versus another. I think it's just number of features for sure.

Some features are more important for one audience, some features are more important for a different audience. Some platforms are meant just for videos. Some platforms are meant for all kinds of media content including videos, images, or even documents like PDFs whatsoever.

Some platforms are targeted enterprise customers, some platforms are more open for startups and for people who just started building. So very different approach and one size doesn't fit all. What is that one thing you see folks still doing these days besides the fact that they're trying to roll their own file management system here for videos, but um, that makes you kinda shake your head a little bit and go, folks, we gotta be a little bit smarter than that.

I think right now everyone is overly optimistic that they can generate everything using LMS and they're firing a load of people expecting the same results or even better results. In some cases it's a bit premature because generated code is not, it's just different from people can write and it should be proofread. Someone should understand how this works.

You cannot just generate, uh, a lot of code. Even if it works, if someone something breaks, someone should understand how to fix it. So it does make things easier.

We do use this in our day-to-day routines, but I think just to be a little slower with changing team structures, getting used to it, getting used, how to develop optimized platforms that are safe and secure because one of the things is that, uh, all the LMS there trade using code that was available. It was open source or some, the LMS had access to some code. But the thing is the best code, most of the best code is proprietary.

So LMS are not trained on the code that was created for secure, scalable platforms. They're based on something that was available on GitHub all the way and it's just very different. So you cannot expect the same quality of the code that use that was used to build Facebook or some Apple software.

You cannot expect to get it, get it written for your software because it wasn't public in the very beginning. So it's not used to train this LOX. Alright folks, you heard it here.

You can of course build anything, but the question is, is what to do with your time and which is more valuable the time you have to allocate to other things or the manual process of managing a bunch of video files that may not provide a whole lot of differentiated value to your endeavor. Hey Igor, thanks for being on the show. Sure.

Thank you Mike. All right. And back to you guys in the studio.