--extra-split should force split scenes at the exact given value, but instead it only splits a veeery little percentage of them at the given value, while all the other scenes get split at that value -1 or -2 or -3 etc.
This causes all modern encoders to break the last gop and fuck up the hierarchical structure, causing extreme losses in efficiency and sometimes even in quality.
Basically, not using av1an would automatically be better.
Example:
I use -x 193 -> av1an splits some of the extra scenes at 193 frames and most of them at 192, 191 and even 190 and 189.
The last gops of all these scenes end up not respecting the hierarchical structure of the encoder and cause massive losses in efficiency.
Extreme example:
I set -x 33 min-scene-len 33 -> av1an splits 20-30 scenes at 33 frames, then the other THOUSAND at 32 and 31 etc.
Why is it only splitting a very few scenes at the given value?
Why is min-scene-len not respected at all? This happens in general too, as there is often one or two scenes that end up being smaller than min-scene-len, but this is not really an issue and may even be better, like when the final chunk of a larger scene is left with less frames than the min-scene-len and respecting the value would mean removing the supposedly way more important keyframe at the next scene change; the real issue is when too many scenes simply don't respect it, like in this case when I'm forcefully splitting all scenes at 33 frames but they instead all get split at 32 or lower. Why aren't the scenes simply split at 33 as requested?
This behavior of not respecting the exact value given at --extra-split obviously destroyes completely the purpose of splitting in chunks and it breaks every single gop.
Do you concur? Am I missing something?
Thanks in advance.
--extra-splitshould force split scenes at the exact given value, but instead it only splits a veeery little percentage of them at the given value, while all the other scenes get split at that value -1 or -2 or -3 etc.This causes all modern encoders to break the last gop and fuck up the hierarchical structure, causing extreme losses in efficiency and sometimes even in quality.
Basically, not using av1an would automatically be better.
Example:
I use
-x 193-> av1an splits some of the extra scenes at 193 frames and most of them at 192, 191 and even 190 and 189.The last gops of all these scenes end up not respecting the hierarchical structure of the encoder and cause massive losses in efficiency.
Extreme example:
I set
-x 33 min-scene-len 33-> av1an splits 20-30 scenes at 33 frames, then the other THOUSAND at 32 and 31 etc.Why is it only splitting a very few scenes at the given value?
Why is
min-scene-lennot respected at all? This happens in general too, as there is often one or two scenes that end up being smaller thanmin-scene-len, but this is not really an issue and may even be better, like when the final chunk of a larger scene is left with less frames than themin-scene-lenand respecting the value would mean removing the supposedly way more important keyframe at the next scene change; the real issue is when too many scenes simply don't respect it, like in this case when I'm forcefully splitting all scenes at 33 frames but they instead all get split at 32 or lower. Why aren't the scenes simply split at 33 as requested?This behavior of not respecting the exact value given at
--extra-splitobviously destroyes completely the purpose of splitting in chunks and it breaks every single gop.Do you concur? Am I missing something?
Thanks in advance.