@dang please delete this self-promotion. I read the entire article and there were no technical details and no value in general. If i come across something like this on Google then I blacklist the domain
I was a lot easier than I thought to build a video delivery system. From a backend point of view:
s3 -> mediaconvert -> s3 -> backblaze -> fastly
gives you a super low cost solution that can scale forever. If you commit to bandwidth for fastly you can get that super cheap.
I've played around with ffmpeg for a few days, and with over 100 videos I couldn't get it to make the video smaller than mediaconvert. That's probably a personal problem, but it makes me wonder how good the ffmpeg encoders actually are. And I was using the original source to compress.
Anyway, good luck to you. The video space is about marketing. It's really crowded, and in the end marketing people don't care how much or how good the solution is...which is why BrightCove and Vimeo exist. They literally suck at everything except marketing. It'll be a great day when they die.
Yeah this article contains no details and sounds self-congratulatory more than anything. As someone currently working with these same technologies for a new and ambitious project this frankly makes me less likely to hair Wistia, not more.
There are plenty of challenges with video transcoding and delivery even today. ffmpeg/libav are still at the core of any transcoding but they are opaque and their integrals are difficult to understand let alone modify. Global CDNs are able to help but you do hit a wall pretty quickly where it goes from cheap to expensive to basically impossible. WebRTC is an emerging technology that is still full of bugs in the spec and implementation. Browsers don’t really handle video decoding in what you might call robust way.
There are plenty of problems to solve in this space. Just saying that you built infrastructure that is “ reliable, scalable, high-quality, secure, and cutting-edge” is not enough to explain how you overcome any challenges. It just sounds like a buzzword salad.
We're hoping to publish some more in-depth posts - but had to start somewhere! We've been deep in ffmpeg for a while, especially around building our in-browser editor + record tools - excited to share more about what we've learned soon.
