keyint=min-keyint vs scenecut=0

[Blue_MiSfit]

Hello!

https://mailman.videolan.org/pipermail/x264-devel/2016-September/011765.html

Did this ever get implemented? (setting keyint to min-keyint and leaving scenecut alone allowing for fixed GOP while still allowing adaptive non-IDR I frame insertion)

It works very well in x265 and makes me sleep better at night knowing the poor encoder doesn't have to code scene changes as P frames :)

With latest zeranoe ffmpeg it doesn't appear to work.

[MasterNobody]

No. And if you read this linked thread fully than you know why. And I never get someone's other comparison that showed keyint=min-keyint is preferred (for H.264 and specifically for x264). So at least in x264 the only way for fixed GOP is disable scenecut (nothing is bad to encode scenecut as P-frame if you already restricted yourself to fixed GOP).

[Blue_MiSfit]

I did read the whole thread - it looked like the benchmarks that were done did not use the proposed patch, so they demonstrated poor results. I could have misinterpreted this though, did you get a different impression?

[MasterNobody]

First of all there was no proposed patch other than simple removing of min-keyint limit. So I tested what was proposed. Also don't take at face value Tom's words in the last post of thread. He is more of marketing/manager person than technical expert or programmer. I hope my words wouldn't offend him. But without other benchmarks that will show advantage of proposed changes that is only words.

[Blue_MiSfit]

Tom's a great guy, I've had some good interactions with him over the years!

I think I did some testing with x265 awhile back on this and found some nice improvements in preventing transient quality dips, but I'll have to review that.

[benwaggoner]

Tom's a great guy, I've had some good interactions with him over the years!
QFT! Tom is extremely technical and directly led or contributed to many important x265 improvements over the years.