DIVX 5.05 Nth Pass Comparison Alpha3

[Xethos]

Comparison Alpha:
http://home.attbi.com/~xethos/index.htm

I will ad more pictures as i encode more video's.

[temporance]

Hey Xethos,

It might be more interesting to take screen shots of still scenes which the eye actually gets to focus on in the real movie. True, the fire is hard to encode, but it also hides artifacts pretty well when played back at x1 speed. So, if a codec spends less bits on the fire and more on still, detailed scenes, then it's doing its job right.

BTW, there's something that looks like a codec error in your 4th JPEG - I wonder how that happened.

temp

[Xethos]

Its not on the pic on my pc...odd.

Yah - i'll add more from more movies. I hapo to have @ least 10 examples.

[Xethos]

If anyone wants to help me, plz email me @
aspicer3 at attbi dot com

Enjoy it.

[jeremymacmull]

what about as well as the multipass conparisons u post what the same scenes would look like with phy on normal b frames etc

but the most important one to post is the original two pass encode what it would look like using original two pass with the same bitrate and settings

JEREMY

[Xethos]

I plan to expand the bitrate's to more than 500 and to include Original 2 pass.

[JohnMK]

I'd love to see a comparison to XviD here too.

[BiaTch 5.0]

I would like to see DivX old 2 pass vs. Nth 4 pass.

IMO

Old 2 pass - less blocking

Nth 4 pass - Sharper

I belive BM should improve Nth pass however which setting to use? testing must be done...

BTW: The high motion blocking issue I was having (& posted about) with 5.05 is the same issue people are now seeing in Nth pass.

[dTb]

Originally posted by BiaTch 5.0
I belive BM should improve Nth pass however which setting to use? testing must be done...

BTW: The high motion blocking issue I was having (& posted about) with 5.05 is the same issue people are now seeing in Nth pass.

Try -0.02, I've read a post by Gej where he said constant bitrate modulation favours low motion a little and that -0.01 or -0.02 will probably result in more consistent quality throughout.
Personally I find the quality of high motion not quite up to my standards so I generally use -0.02, this might help solve the issue you have with high motion.

[CaptainCarrot]

Uhm, I have a little philosophical question about the usefulness of this comparison:
DivX changes the bitrate distribution from pass to pass to optimize the overall quality. So no matter what encodes you compare, be it 2pass or 6pass, each will have frames that look better and frames that look worse. So comparing indivdual frames is more or less pointless, since the codec behind it is the same, it only gets more or less bits to spend, which then costs or pays for other scenes.
To really do a comparison that allows you to judge overall quality you'd have to watch the movie and point out the (let's say 10) most annoying scenes, or count the annoying scenes or something like that. The optimal version would be the one that keeps it's quality level constant over the whole movie whitout any scene that shows notably more artefacts than the average.
I know that this would be a really time consuming test scenario as you had to watch all versions of the movie completely, and you'd have to be pretty concentrated all the time. But honestly, every other
quality-comparison between different passes is pointless.

[BiaTch 5.0]

Originally posted by dTb
Try -0.02, I've read a post by Gej where he said constant bitrate modulation favours low motion a little and that -0.01 or -0.02 will probably result in more consistent quality throughout.
Personally I find the quality of high motion not quite up to my standards so I generally use -0.02, this might help solve the issue you have with high motion.
Thanx, I'll try -0.02.

[SeeMoreDigital]

Your tests at 500kbps are very low for DivX.

I've done some mixed low & fast action tests using DivX at 600kbps, (using a 720 x 576 pixel source) and that's about as low as I would go.

However, I've found the longer the source file (60mins plus) the better the multi pass encode becomes.

I've also found that anything over 6 passes is not worth the extra effort.

But interesting all the same.