Getting video online before your competitors is more important than ever, for better or for worse. Launching today at the Disrupt Battlefield, QuickFire.TV claims to transcode video ten times faster than similar services, providing news and sports organizations videos encoded for playback on a range of devices minutes after it’s uploaded from source footage.
QuickFire’s speed is the result of the startup building a complete stack for its service, including custom hardware and software designed specifically to reduce transcoding times. Each rack in the servers that make up its hardware platform (which the company calls T-Video) contains a custom motherboard built to accommodate 11 quad-core Intel Core i7 processors. Each of those processors work in unison, with work distributed by a software layer QuickFire calls V-Fabric. Furthermore, each rack in the server is tied together with another layer known as Q-Data.
Lee demonstrated the platform on-stage by transcoding “Big Buck Bunny,” which he called “The Kardashians of video editing — it’s everywhere.” In seconds, he encoded 50 minutes of video for 20 different platforms.
Considering the versatility of graphics chips from the likes of Nvidia in tasks that can be processed across many cores, I expected QuickFire’s servers to heavily rely on large clusters of them. But before today’s presentation, QuickFire TV CEO Craig Lee told me that the company’s servers don’t use any graphics processors except those built into Intel’s high-end CPUs. That’s not the case in applications I’m familiar with (video games and video editing), but Lee was adamant: “Intel is a force to be reckoned with in the GPU space.”
In addition to providing a complete stack on its end, QuickFire offers a set of APIs with documentation and sample code so video production teams can put together their own workflows and custom tools for uploading videos and checking transcoding progress. While telling QuickFire which profiles (frame rate, resolution, color and audio settings for different platforms) to generate still has to be done with a quick call or email chat, it only has to be done once.
After a free evaluation period, QuickFire charges its customers a base fee per minute of output: $0.10 per minute for standard-def video (for streaming to viewers with low bandwidth or on mobile) and $0.20 per minute for HD. Lee tells me production teams that regularly transcode large video segments can also negotiate discounts.
After their presentation, the Battlefield judges asked Lee a few questions covering things he didn’t directly touch on. Here’s the gist of what they asked about and Lee’s responses (they do these things rapid-fire):
Is anyone using this in production?
No, not in production.
How much can you guys process at any given moment? Is there a limit to the users you can handle at once?
The key is to find the customers willing to pay a premium for speed — we can build up on any cloud.
Most of the stuff that I post to social media is a few seconds long. Do regular users really need this?
We think that the current upload times are too long for younger users — they simply don’t want to wait a few hours or overnight to see their videos on YouTube.
Have you built in systems for uploading to other platforms? Can I set my transcode jobs to immediately go to, say, YouTube, Vimeo, and so on?
We’re part of the regular video workflow — just tell us where it needs to go and we’ll transcode it. We don’t have a content management system, we only get your videos ready for wherever they need to be.