api.video made some waves recently when they announced zero-cost video encoding. As detailed by company CEO Cédric Montet, “Encoding videos will be absolutely free for all video qualities (1080p or 4K) at api.video – no hidden calculations, footer notes or tricks.”
FIGURE 1. Free video encoding
Using specialized ASIC encoding hardware from NETINT enabled api.video to 10x their transcoding capacity without sacrificing quality. In the post “Free video encoding: Let’s talk numbers,” by Reich Jean-Baptiste and Arushi Gupta details for this new business model were shared including the quality analysis.
Encoding Costs Down 99.33%
The secret to achieving this incredible cost savings was the decision to deploy their own infrastructure rather than build on a cloud provider. The authors commented, “With VPUs [Video Processing Units], we are able to save power consumption costs since the power consumption cost of VPUs is one order of magnitude lower than CPUs and GPUs. It is also easily possible to pack a lot of them in a single server. This increases their density and, reduces the number of servers.”
The authors then focused on fully loaded encoding cost, considering loaded server cost (CAPEX) and electricity cost (TCO). After running the math, they found that “If on average, this server is loaded at 50% (12 hours a day), the cost per minute would be 1/150 of our previous price.” This translates to a 99.33% cost reduction simply by switching to NETINT’s ASIC-based VPUs. If your encoding costs dropped 99.33%, I’m sure you could be pretty generous with your customers, too, whether internal or external.
With No Reduction in Video Quality
The authors then focused on video quality, noting that “Hardware encoding is always associated with quality losses in the head of many video engineers. But that was probably the case decades ago. Not anymore.” To prove their case, the authors compared output from the ASIC-powered api.video encoder with files produced by FFmpeg using x254 and the Medium preset (Table 1).
We’re big fans of Netflix’s VMAF metric and take pride in outpointing FFmpeg there, though we admit that a .59-point difference would be invisible to all viewers. Reverse ranking but the same result with PSNR and SSIM; though FFmpeg outpointed the api.video file, the difference was irrelevant. The bottom line is that api.video output would be visually indistinguishable from FFmpeg.
TABLE 1. As measured by three metrics, api.video output would be visually indistinguishabe from FFmpeg using x264 and the Medium preset.
In sum, api.video’s decision to offer free video encoding showcases the company’s commitment to innovation and customer satisfaction. By harnessing cutting-edge VPUs and building their own infrastructure, api.video reduced encoding costs by 99.33% without compromising video quality. At NETINT, we’re proud to have helped enable this transformation, and we’re excited about the possibilities our encoding VPUs might hold for your future endeavors.