Do we need Anti-shit for DivX 5.03 now??!!

[DJ Bobo]

You won't believe it until you see it!

DivX 5.03 is producing shit, for the first time since DivX 3.11 SBC!!
I encoded with the normal version in 2-pass mode (well, that nth pass thing), and I got shit frames for 2 seconds (in a 90 minute movie).

You won't see this very often, so check it out: (I still don't believe my eyes!!)

[The Edge]

Can you post a screenshot?

Edge

[jggimi]

He did, and I validated it. 67KB.

[BOB]

Nasty,

Thus far i've done quite a few cartoon episodes using 5.0.3, mostly just trying nth pass etc
So far i've had nothing like that happen, i've started doing a proper long encode of a film so I guess if something like that happens it will happen on the film.
What settings were you using and what method were you using to rip?
ie. Through vfapi or something like xmpeg/flashmpeg etc.

BOB

[DJ Bobo]

My usual way: DVD2AVI - AviSynth - VirtualDub
DivX normal version, all on default settings, no psycho, nth pass (2-pass) @820kbps.
Nothing special. (the problem has nothing to do with me or my methods anyway :D)

Seems like DivX Networks made a new bug ;)

I think it's time to return to the original 2-pass...

[Slogra]

Where is the shit? I thought shit was little blocks with inverted colors???

[temporance]

That's not shit, although I agree it looks crap. It looks like the codec is running out of bits and dropping the bitrate. I've had something like this happen when using very low max bitrate. What is the setting of your max bitrate in your 2nd pass, DJ Bobo? Does this happen anywhere else in your movie? What does the 3rd pass look like?

[Morbo]

IMHO it is crap.

To fit on two cds with AC3,the clip had to be encoded at 700kbps.
Scene was a flashy bar room scene in Fire Walk With Me
and DIVX just doesnt do it justice at any bit rate for me.

The only way I can see DIVX nth pass being useful,is if
Gknot could keep the best frames from all the passes,instead of
always just making another pass.

Cheers!!

PS : Still think DIVX is a good encoder.try the new version(503)
on the opening credits of Austin Powers(1st one).It Chokes.

[yingx2]

The only way I can see DIVX nth pass being useful,is if Gknot could keep the best frames from all the passes,instead of always just making another pass.

Technical issues aside, I think the rate control system would totally break down by doing so -- you will always get oversized files.

(Best frames = frames encoded with lowest quant value = largest frames) A collection of largest frames makes an oversized file.

[DJ Bobo]

If this is not shit, what is it then?! (always picking on non important terms ¬.¬)
I never saw somehting like this before, other than in DivX 3.11 SBC encodings. So what was the antishit function for anyway?! ¬.¬
inverted or not inverted, it's all shit!

Anyway, home theater profile was selected, which has 4000kbps max.
No 3rd pass was done (and never intended or intend in my future encodings to do a 3rd one! it just takes too long already)

But this has nothing to do with max bitrate, this is a case where the codec allocates far less bits than needed. This happens as stated above only for those 2 seconds in that (90 minute) movie.

I'm back to the original 2-pass already (this is not a thread to ask how to solve this, I just wanted to share it with you, because you won't see such a thing every day. There is nothing one can do about it anyway, when it happens, you'll have to replace the shit part with a fresh one)

[angelyote]

I saw something similar to this when doing a 2cd encode of Castle of Cagliostro. For the most part the encode looked ok but in darker scenes it was complete crap. Huge blocks all over. It looked like I was encoding at 600kbps instead of 1800.

My solution to this was to rollback to DivX 5.02. Problem solved and I get my quantizer tweak back without needing to hack the registry.

Dave

[BOB]

Uhoh...

I've just experienced something similar..
It could have been my dvd drive though, it is getting on a bit now.. on the plus side i'm getting a Sony DRU500A delivered tomorrow :D
Something could have messed up when it decss'd it to the harddrive, suppose I should play the entire decss'd vobs on the harddrive to check.

DivX settings used were multipass using 10 passes, keyframes every 250, psychovisual enhancements at normal, bitrate was about 1000 iirc, with gmc and bi-frames enabled.

See pic

BOB

http://home.clara.net/mccasey/weird_dilbert.jpg

[Kheffer]

"psychovisual enhancements" sometimes produces this kind of artifacts. Try disabling it.

[Suikun]

If you watch that clip with ffdshow and look at the quantizers, you'll notice that DivX used the highest possible quantizer of 31 for that scene, thus making it, of course, impossible to watch. Try encoding a movie at quant 31 and you'll see...
My guess is that it's a bug in DivX's bitrate curve treatment (some overflow maybe...) or a corrupted statsfile.

[BOB]

Funny thing is i've done a ton load of multipass encodes now using divx 5.0.3 with psychovisual enhancements at normal, had no problems until this one appeared.
Anyway, got my firmware updated Sony DRU500A from Bigpockets :D
Woo!
WOO!
WOOOOOO!!!!

I'm just waiting for this restarted Dilbert encode to finish, check it to see if its the same as the previous one and then install my brand spanking new dvd writer. :D

WOO!

BOB

[Krack]

I've got crap with two different videos too !! :( (no psy original 2 pass 600k/s)
It seems that it append in dark scenes. I will reencode these two to see if it may produce the same shit in the same place ...

rectif: It seems thet the artifacts were already on one of the sources (John Doe 1x05). It may be the same with the other one.