Mastodawn

I'm glad that *somebody* is working on a video codec that's optimised for stereoscopic video, but wow I hate the way that Apple has made MV-HEVC quasi-proprietary.

Show thread

Misofist 2d ago

Ok even compiling ffmpeg and x265 from source, I am unable to successfully transcode stereoscopic video on a round trip through MV-HEVC. The encode *seems* to work (with some warnings), but I have no way to test it. Decoding fails with a bunch of errors.

Every single person online I can find who got it to work was using macOS. It seems it might only be possible on Apple hardware?

NvEnc allegedly has support but with zero documentation. Unclear even what GPUs support it.

Show thread

Kevin Karhan

1d ago

@Misofist that is by design!

Because the people behind #ValueRemoving & #RentSeeking proprietary tech like #HEVC want to shaft consumers and manufacturers for #licensing money, because that's how their business ROI's itself!
I just wished #AV1 would have feature parity in that regard…

Show thread

Misofist 1d ago

@kkarhan Yeah. Funnily enough only previous attemp I could find at stereoscopic video encoding was MVC, which is an extension to h264.

It'd be so cool if AV1 pursued this! But it's still a pretty niche use case so I'm sure it's not a priority for them.

Show thread

Kevin Karhan

1d ago

@Misofist yeah, espechally since we know how #3D-#BluRay solved that withctheir over/under frames + black blanking interval to make itBwork.

I guess it was like 1920×2200p24 or something…

Show thread

Misofist 1d ago

@kkarhan Yeah I mean you can do that but it doesn't play nice with modern compression algorithms. Side-by-side (current industry standard) is better, but very wasteful due to not taking advantage of the similarities between the two sides.

Show thread

Kevin Karhan

@Misofist granted, 3D + 360° video files are the trickiest to compress at the peak where lossy compression isn't noticeable.

The best results I've seen were two distinct video per eye in a compressed MKV container, but sadly I don't have access to that file anymore.
The only reason they don't do something similar with 3D cinemas is because there the setups are designed with throughput in mind and not computational complexity.