The Case for Shot-Based Encoding
See more videos like this on StreamingMedia.com.
Learn more about per-title encoding at Streaming Media East Connect 2021.
Read the complete transcript of this clip:
Jan Ozer: In 2018, Netflix invented shot-based encoding. And what you do there is, you divide each video into shots and encode each shot separately. You parse the video, you split its scene changes, then you encode each scene separately. So here we see the parsing. We see I-frames at the different scene changes here. We see 1080p, 720p, 480p, and we're going to encode each one of these scenes separately in each rung in the encoding ladder.
And why is this beneficial? It's the most logical approach for applying compression parameters. Before shot-based encoding, we would encode two-second chunks or three-second chunks, which makes no sense because that doesn't align with the content in any meaningful way. And because you're changing your encoding parameters at scene changes, it's going to be less noticeable than if you change the key encoding parameter, say, halfway through a scene.
Obviously, you see that because you're using I-frames in each rung and the encoding ladder at the same location, you're going to preserve ABR switching algorithms so you can switch from stream to stream without any problem.
The other reason shot-based encoding makes a lot of sense is that the benefits are very significant. So with X.264, Netflix saw 28% savings overall as measured by PSNR and VMAF, 37% for VP9 and 32% for HEVC. So obviously, per-title or shot-based encoding are not codec-dependent.
The key benefits are, it's more efficient encoding. The key weaknesses are, you can't have it, because Netflix invented it. And as far as I know, there are no commercial implementations of this, though I'm sure they're coming. And the other weakness is that it's more complex than two-second GoPs and six-second segment sizes. You've got to have different GoP.
Streaming Learning Center's Jan Ozer explains the importance of adjusting the rungs on your encoding ladder in this clip from his presentation at Streaming Media Connect 2021.