Seems to me that 5.2.1 retains less detail than 5.1.1...

[Fr4nz]

Hi guys! I've encoded 3-4 movies so far with 5.2.1 and it seems to me that it retains LESS detail than 5.1.1.

Usually I use these settings:

- 2 passes
- Standard setting (for all the 2 passes)
- H.263 matrix for all the 2 passes in 5.2.1(in fact you canchoose the quantization matrix in 5.2.1...I don't know which matrix is used in 5.1.1)
- Keyframe interval at 250 (I work on PAL material).

I read somewhere that H.263 matrix gives LESS detail than the MPEG matrix, so it is suggested for low bitrates. But I use bitrates around 1.300, so they're not SO low...maybe I should use a combination of MPEG/H.263 over the 2 passes in order to retain more detail?

Unluckily I've read in this thread http://forum.doom9.org/showthread.php?s=&threadid=80347 that something is wrong with the MPEG matrix...

Thanks for any answer!
Fr4nz

[stephanV]

Originally posted by Fr4nz
Hi guys! I've encoded 3-4 movies so far with 5.2.1 and it seems to me that it retains LESS detail than 5.1.1.
Could you perhaps give some screen shots?


I read somewhere that H.263 matrix gives LESS detail than the MPEG matrix, so it is suggested for low bitrates. But I use bitrates around 1.300, so they're not SO low...

Not low, but not extremely high either, I'd still use H.263 if i were you.


maybe I should use a combination of MPEG/H.263 over the 2 passes in order to retain more detail?

No that wouldnt make sense, and probably isnt even allowed.


Unluckily I've read in this thread http://forum.doom9.org/showthread.php?s=&threadid=80347 that something is wrong with the MPEG matrix...

Those are mere assumptions... perhaps DivX people can shed some light on that.

[Fr4nz]

Originally posted by stephanV
Could you perhaps give some screen shots?


hey man, you know that taking screenshot from a movie can make you think wrong things... :) In fact you should see the movie during playback in order to judge correctly. If there's someone who want to judge the differences I can try to put online 2 video-clips, one made with 5.1.1 and the another one made with 5.2.1. Let me know...

PS: Stephan could you tell me which matrix DivX 5.1.1 uses during its encodings??

PS2: I forgot to say before that I used the new b-frames feature "adaptive multiple consecutive" when I encoded with 5.2.1.

[stephanV]

Originally posted by Fr4nz
hey man, you know that taking screenshot from a movie can make you think wrong things... :)
True, but if things are that obvious a screenshot would suffice. :p


PS: Stephan could you tell me which matrix DivX 5.1.1 uses during its encodings??

H.263


PS2: I forgot to say before that I used the new b-frames feature "adaptive multiple consecutive" when I encoded with 5.2.1.
Hmm, I'm not sure how that would affect things... did you use B-frames with DivX 5.1.1?

[Fr4nz]

Originally posted by stephanV
True, but if things are that obvious a screenshot would suffice. :p
Yes you're right :) When I'll notice something of REALLY evident I'll post it.

Hmm, I'm not sure how that would affect things... did you use B-frames with DivX 5.1.1?
Yes I did.

Anyway, with these settings, using both 5.1.1 and 5.2.1:

- 2 passes @ 1300kbps
- Standard mode
- H.263 matrix (used as default by 5.1.1, selected via menu with 5.2.1)
- B Frames activated
- 250 Max keyframe interval
- Psicovisual enhancements deactivated

can we say that 5.2.1 should produce, at least, encodings of EQUAL quality if compared to 5.1.1?

[DigitAl56K]

Originally posted by Fr4nz
can we say that 5.2.1 should produce, at least, encodings of EQUAL quality if compared to 5.1.1? [/B]

Yes it should. 5.1.1 used H.263 matrix also.

[Sharktooth]

Adaptive multiple B-Frames is your "problem".
Use adaptive single consecutive or the old b-frames behaviour using the CLI (dont remember how... but if you use the forum search feature you will find how to do it).

[Fr4nz]

Originally posted by Sharktooth
Adaptive multiple B-Frames is your "problem".
Use adaptive single consecutive or the old b-frames behaviour using the CLI (dont remember how... but if you use the forum search feature you will find how to do it).

Ok, but why you suggest to avoid this new feature? Does it have something to do with the "less detail" problem I've noticed?

Anyway, I hope at least that adaptive single consecutive mode is as good as the old b-frames behaviour....

[Sharktooth]

Multiple b-frames sometimes reduce the movie quality (a lot) and i dont like how the new divx adaptive mode chooses "when" to place a bframe(s) or not.

[Fr4nz]

Originally posted by Sharktooth
Multiple b-frames sometimes reduce the movie quality (a lot) and i dont like how the new divx adaptive mode chooses "when" to place a bframe(s) or not.

Mmmhh I understand. At least is it safe to use "Adaptive single consecutive"?

Or is it better to return to divx 5.1.1?

[Sharktooth]

I still prefer the old method (u can still use it with 5.2.1 by setting a parameter in the CLI - sorry i don't remember what one) but "adaptive single consecutive" should be safe.

EDIT: Fixed typos