Jump to content

Netflix Describes New Per-Title Encode Optimization


Recommended Posts

Being one of the largest streaming platforms, serving people with various devices and Internet connections, Netflix has a lot to consider when it encodes titles. Back in 2010 when it started using h.264/AVC to encode its catalog, the company constructed a bitrate ladder for all of its videos, which based the video bitrate on the output resolution. For most content this ladder works fine, but because some content needs more or less data to be of high quality, Netflix has developed a per-title encoding approach.

Instead of applying a fixed bitrate ladder to every video, this approach constructs the ladder for each video. This means that relatively simple videos, like cartoons, will use lower bitrates than more complex videos, at a given resolution. In one example, this new approach found that a cartoon could be encoded at 1540 Kbps for 1080p resolution, while the older bitrate ladder was using 1750 Kbps for 480p. While the 480p video may have a higher bitrate, upscaling the video introduces in visible artifacting.

When it comes to video encoding, they are diminishing returns at high bitrates, as well as a precipitous drop at lower bitrates, with the optimal bitrate somewhere in the middle. Instead of using a fixed approach, this per-title optimization is meant to find that optimal point for each video, so Netflix can provide the best quality without unnecessarily high bitrates. You can check out the Netflix blog post for the graphs that demonstrate this point, and while it is a long post, it does a decent job explaining it.

Source: Netflix

Back to original news post

Share this post

Link to post
Share on other sites


  • Create New...