flebber
31st December 2008, 13:59
First I haven't used Xvid extensively I have largely used x264. I am using megui which has he 1.2.1 + vaq version of xvid.
I did two encodes expecting CQ profile to be best in both size and quality, however to me the 2pass version has had a better result, though it also does have a larger file size.
here are two frames
first 2P 1000
http://farm4.static.flickr.com/3242/3153651804_4daea95b2e_o.jpg
http://farm4.static.flickr.com/3237/3153651792_992e63f350_o.jpg
and next the CQ2
http://farm4.static.flickr.com/3237/3153651792_992e63f350_o.jpg
http://farm4.static.flickr.com/3242/3153651804_4daea95b2e_o.jpg
here are the profiles
2P 1000 -
program -i "input" -pass2 ".stats" -bitrate 1000 -kboost 100 -chigh 30 -clow 15 -overhead 0 -turbo -max_key_interval 250 -nopacked -vhqmode 4 -qmatrix "C:\Program Files\megui\extra\eqm_v3lr.xcm" -closed_gop -lumimasking -notrellis -imin 1 -pmin 1 -max_bframes 1 -bvhq -bquant_ratio 162 -bquant_offset 0 -bmin 1 -threads 1 -o "output"
CQ2
program -i "input" -single -cq 1.5 -smoother 0 -max_key_interval 250 -nopacked -vhqmode 4 -qpel -closed_gop -lumimasking -max_bframes 1 -bvhq -bquant_ratio 162 -bquant_offset 0 -threads 1 -o "output"
Is this a usual result for xvid ? Can CQ 2 be improved any further, without a big increase in file size? As the Xvid 2P 1000 finishes 20% bigger than the CQ2, in improving CQ2 I wouldn't want to exceed this % size increase.
I am used of getting 1 pass CRF results in x264 being better than 2 Pass when file size is of no concern.
PS. I have found an extra difference between encodes the 2P 1000 uses - C:\Program Files\megui\extra\eqm_v3lr.xcm - this quantization, whilst CQ2 uses h263.
Just for reference here is the same frame encoded with x264 on a 2 Pass Its final file size was minimally different to the final size of xvid CQ2.
the x264 profile.
http://farm4.static.flickr.com/3196/3153690674_e0db49c641_o.jpg
program --pass 2 --bitrate 1050 --stats ".stats" --level 5.1 --ref 5 --mixed-refs --no-fast-pskip --bframes 3 --b-pyramid --weightb --direct auto --deblock -1:-1 --trellis 1 --partitions p8x8,b8x8,i4x4,i8x8 --8x8dct --me umh --merange 24 --threads auto --thread-input --aq-strength 1.2 --progress --no-dct-decimate --no-psnr --no-ssim --nr 300 --output "output" "input"
PpS. This just for a bit of fun.:)
I did two encodes expecting CQ profile to be best in both size and quality, however to me the 2pass version has had a better result, though it also does have a larger file size.
here are two frames
first 2P 1000
http://farm4.static.flickr.com/3242/3153651804_4daea95b2e_o.jpg
http://farm4.static.flickr.com/3237/3153651792_992e63f350_o.jpg
and next the CQ2
http://farm4.static.flickr.com/3237/3153651792_992e63f350_o.jpg
http://farm4.static.flickr.com/3242/3153651804_4daea95b2e_o.jpg
here are the profiles
2P 1000 -
program -i "input" -pass2 ".stats" -bitrate 1000 -kboost 100 -chigh 30 -clow 15 -overhead 0 -turbo -max_key_interval 250 -nopacked -vhqmode 4 -qmatrix "C:\Program Files\megui\extra\eqm_v3lr.xcm" -closed_gop -lumimasking -notrellis -imin 1 -pmin 1 -max_bframes 1 -bvhq -bquant_ratio 162 -bquant_offset 0 -bmin 1 -threads 1 -o "output"
CQ2
program -i "input" -single -cq 1.5 -smoother 0 -max_key_interval 250 -nopacked -vhqmode 4 -qpel -closed_gop -lumimasking -max_bframes 1 -bvhq -bquant_ratio 162 -bquant_offset 0 -threads 1 -o "output"
Is this a usual result for xvid ? Can CQ 2 be improved any further, without a big increase in file size? As the Xvid 2P 1000 finishes 20% bigger than the CQ2, in improving CQ2 I wouldn't want to exceed this % size increase.
I am used of getting 1 pass CRF results in x264 being better than 2 Pass when file size is of no concern.
PS. I have found an extra difference between encodes the 2P 1000 uses - C:\Program Files\megui\extra\eqm_v3lr.xcm - this quantization, whilst CQ2 uses h263.
Just for reference here is the same frame encoded with x264 on a 2 Pass Its final file size was minimally different to the final size of xvid CQ2.
the x264 profile.
http://farm4.static.flickr.com/3196/3153690674_e0db49c641_o.jpg
program --pass 2 --bitrate 1050 --stats ".stats" --level 5.1 --ref 5 --mixed-refs --no-fast-pskip --bframes 3 --b-pyramid --weightb --direct auto --deblock -1:-1 --trellis 1 --partitions p8x8,b8x8,i4x4,i8x8 --8x8dct --me umh --merange 24 --threads auto --thread-input --aq-strength 1.2 --progress --no-dct-decimate --no-psnr --no-ssim --nr 300 --output "output" "input"
PpS. This just for a bit of fun.:)