DivX5.1 2nd Test! forget the 1st....

[SpikeSpiegel]

Good news!
I've red some posts of DivX.com forum and I found out that my 1st tests weren't correct:
The performance/quality ratio can be used freely in multipass, so u can do the 1st in "standard" and the nth in "slow" mode (and vice-versa), BUT NOT if mvinfo file is enabled!
For all tests I used "write/read mvinfo", that's why I had "noisy" videos and the speed in nth pass with slowest mode was so fast!
Moreover it seems that standard mode of divx5.1 corresponds to the improved version of “slowest mode” of 5.05, so, even in this mode, it should be better than divx5.05.

So the new test’s results are:
Standard mode:
1)Speed (same criteria: no options enabled (except bi-directional frames), compared to divX 5.05)
Without mvinfo:
-10/20% 1st pass, -10% nth pass
With mvinfo:
-10/30% 1st pass, +10/20% nth pass (at least is almost the same of 5.05)

[SpikeSpiegel]

2)Quality
Yes, actually I can confirm that the motion search/data allocation precision is really improved, I saw macroblocks only using extremely low bitrate values.
Even encoding anime videos, the noise is barely noticeable, much better than divX5.05.
Comparing it to Xvid, the most differences are in high motion scenes: with divx there’s less noise and no macrob.

Slow and Slowest mode:
1)Quality (the good news first…)
Noise? What does this enigmatic word mean?
Simply perfect. I encoded the same anime trailer (res=512*384, 1000kbps, 25fps), but this time the noise vanished…it was like using convolution3D!
Then I re-encoded a normal movie (obviously only 1little part of a chapter): no noise, more crispness.

2)Speed…speed?! How can I call this “speed”?!:
Slow
1st pass: 1.7FPS
nth pass: 2.4FPS

Slowest
1st pass: 1.2FPS
nth pass: 1.9FPS
….usually with DivX5.05 fps=10!

[SpikeSpiegel]

Conclusions:
If u need perfect quality use Slow mode (I haven’t noticed any difference with slowest)
If u think that life is too short to encode one movie in one century, then the standard mode is what u need: it’s fast as divX5.05’s slowest mode (the default one), but I think that produces better quality

Let me know what do u think about it!

[nithro]

Thanks for your time and help.

If u think that life is too short to encode one movie in one century, then the standard mode is what u need: it’s fast as divX5.05’s slowest mode (the default one), but I think that produces better quality

So you think in slow mode you get the best quality.
But using the both 1 and 2 pass in slow?; or you can use the 1st in standard and the 2nd in slow and the quality it's the same?
Thanks again

[StarRide]

Yeah as stated in the divx forums, u can use standard for 1st pass and then Slow/Slowest for the second pass, BUT u must turn off using MV (motion vector) file on the second pass.

[nithro]

Yes, but the quality it's the same?.

u can use standard for 1st pass and then Slow/Slowest for the second pass, BUT u must turn off using MV (motion vector) file on the second pass.

You got the same quality if you use slow in both passes?

[SpikeSpiegel]

In a DivX.com forum's post

http://forums.divx.com/viewtopic.php?topic=53403&forum=5

is stated that the difference of quality between 1&2pass in slowest and 1(stand)&2(slowest) is relatively (considering the time) really little.

I think that 1(slow)&nth(stand) should be better because the first pass (when the movie is analysed) is more important

Anyway the"plus" of slowest is that uses an advanced "rate distorsion control" or somthing like that, so distributes the bitrate better, BUT if if use high bitrate values the difference could be not noticeable, so I suggest u to do some tests

[nithro]

I try the both passes in slow, an the 1st take 20 hours and it's running right now.
This is the only problem the speed. I guess the 2nd pass it's going to take 20 hours too.
Thanks for your help.

[SpikeSpiegel]

I tested the Psychovisual Enc and, seeing the results, this is the 1st thing I said:
IS IT A JOKE?! (....in Italian, obviously)

It's slower (slow: -33%; fast: -17%) and TOTALLY USELESS!
If u read the last Guide u'll know that Psychov. SHOULD allocate more data where it's needed reducing the noise where it's more visible (especially with anime movies and in fast mode) and SHOULD increase the number of details in the more focused areas.
The only thing that increased are NOISE AND MACROBLOCKS!
First I encoded an anime video with and without psycho(fast): with psyco the noise increased except in some pink (only bright tone) coloured areas (faces)...where the noise is just a little more thin, but more annoying.
Then I did the same with a normal movie and using the slow mode, and two are the possibilities:
1) even in slow mode, psychov. creates noise&macroblocks
2) the actors had a terrible syndrome of acne!

The problem is that now psychov. is selected by DEFAULT!

[ivwshane]

Damn!! 20 hours to encode a video? Ouch!

So if I do a two pass and set the first pass to slow (or slowest) and the second pass to standard do I need to disable the mvinfo for the first one (or which ever one has the "slowest" setting or will I always disable the second passes mvinfo)?

[SpikeSpiegel]

@ ivwshane
You can use Mvinfo only if 1st and nth pass are in the same mode.
Enabling mvinfo only in one of the passes is useless

@ everyone
Don't use slow and slowest mode!
Even in standard mode u can have better quality than divX5.05 (in slowest), anyway the difference of quality between stand&slow mode is really little (comparing it to the time u lose)!