Quote:
Originally Posted by KillerZero
1)How do you know it is PCscale? use script with just AVCSource and ColorYUV(analyze=true) and check loose min luma.
|
Did that, and thats how i know.
Quote:
Originally Posted by KillerZero
1)2) why are you converting it to YUY2? why don't you leave it in YV12?
|
Converting to MPEG2 using CCE (to be precise: Replace the crappy PQ of the video-track of one of my backups with a new and better one)
Quote:
Originally Posted by KillerZero
1)3) Why fft3D? Denoising ?
|
Yes, to help compression + it is needed (using low sigma-values, after resizing from HD to SD).
Quote:
Originally Posted by KillerZero
5) I don't know AVCSource and I cant google any relevant page, but if deblocking=false disables deblocking even if clip uses in-loop deblocking, this can lead to catastrophic results.
|
Read all about it on doom9's forum, only not here but here:
http://forum.doom9.org/showthread.php?t=122598
No catastrophic results known, and it is not relevant in this case.
In other words: I know what i'm doing, so above tips are known to me.
Quote:
Originally Posted by KillerZero
1)4)
I think there won't be any significant difference, but maybe would be better put it after FFT3D, so it will work with more colors.
|
Thx for this answer; But how do you think it works best on more colors ?
My hypothesis for option 1 was was because i do not know if luma getss "cut off" or otherwise broken when i use ConvertToYUY2() on PC-scale input.
If the conversion wont give those problems, than i wanted to use option 2 because any rounding-off errors or other noise (if any) that comes with the conversion gets corrected, and rest of the script can use the as much luma/chroma information i can give it (The FFT is needed anyway).
In the end i want to lose as less color/luma info as possible before giving it to CCE. In other words: The higher the quality the better
(=Didee-like encoding quality-standards
)