View Full Version : Divx5 & 1400mb encoding problem
Excelsior
14th March 2002, 21:10
First of all I'm quite new to this encoding stuff, I have only encoded a couple of movies. If I've understood it corectly that if I put a high bitrate, say ~1400 the movie should get bigger ~2cds. I've tried two times to set the bitrate to 1400 & 2000 in 2nd pass but the movie always ends up 1gb. What's wrong?
theReal
14th March 2002, 22:52
First it depends on the length of your movie and on the bitrate, but it also depends on how compressible this movie is and what resolution you are useing.
First, you need a bitrate calculator to see what bitrate would result in what size for the specific length of your movie. If it doesn't reach the desired size then (which is not the case usually, but it might happen), it means the quality is maxed out and you must increase the resolution and maybe use a sharper resizing method to be able to reach the desired size.
EvilFoo
15th March 2002, 03:41
theReal: Hmm... I have a ~56 minute DVD clip for my first CD. With a resolution around 672x288 and 192 VBR kbit/s audio, it only reaches around 500 mb with a bitrate of ~1600 (which was given to me by Gknot). I can increase the resolution by one more notch into the 700's, but after that red-colored values start popping up.
What do you mean by "sharper resizing method"? How do I do that?
Thx.
theReal
15th March 2002, 04:34
When you are using Gknot, you can choose between bilinear, soft bicubic, neutral bicubic and sharp bicubic resizing. In your case, I'd use sharp bicubic resizing, which is the the most exact resizing method, but also makes the file less compressible.
56 Minutes isn't long and I guess the maximum quality was reached - that is, if you set the codec to encode at 100% quality instead of a certain bitrate, the file will also be 500MB.
However, this might also be a problem of Gknot which was not made for use with DivX5. Maybe it just doesn't set the codec right. When you want to use Gknot with Divx 5 you must take care of a lot of settings manually (I haven't yet tried divx5 with a DVD, only with avi files, but I guess you must manually load the Gknot avisynth scripts in Virtualdub so you can adjust the divx5 encoder yourself)
EvilFoo
15th March 2002, 05:51
Ah well yeah, I always use sharp bicubic... Okay I've gotten the res up to 704x304 making the bits value .303. If doing these extra changes won't make up for that 700 filesize, I might as well include subtitles and different languages!
Here's a quick question, what is the real purpose of doing those compression checks (leading to those values under the bits/pixel*(frame))?
Excelsior
15th March 2002, 20:45
Thank you for your answers the movie was "The mummy", maybe I reached maxium quality with my chosen resolution, because the movie didn't get any bigger no matter what bitrate I chose. I'll see what happenes if I chose a higher resolution and sharp bicubic resizing.
dragoman
18th March 2002, 11:56
Hi,
The movie has probably reached what I call the "saturation point". This means that every frame has been encoded with a quant of 2 (100% quality) and that means that you cannot get the filesize any bigger no matter what, it is the limit.
Xvid allows you to encode with a quant of 1, which might increase your filesize a little, but it prob wouldn't be worth it.
Sorry, man, try adding some special features, ac3 audio, extra audio tracks, whatever to get your filesize up to 1.4gig
dragoman
Franky
19th March 2002, 00:41
>The movie has probably reached what I call the "saturation point".
>This means that every frame has been encoded with a quant of 2 (100%
>quality) and that means that you cannot get the filesize any bigger no
>matter what, it is the limit.
Actually, I don't think the codec actually encodes all frames with quant 2, it encodes every frame at the highest quant it thinks is the best for that frame. For example, a totally black frame is much more compressable than a high action scene, so that particular frame doesn't need a quant higher than, say, 12. This is the reason DivX 5 produces smaller filesizes than older DivX versions. If the older versions gets extra bits available to encode a frame, it automatically uses a higher quant level, whereas DivX 5 just uses the quant it thinks is best for the frame.
karloz1
19th March 2002, 01:46
If divx5 is like other two pass methods, then it doesn't matter what the second pass bitrate is. If you set the bitrate to 1400 on the first pass, the second pass can't be raised higher or lower, it'll just be 1400 as well. The second pass is setting the bitrate from the divx log file. If it's following a log file for 1400 bitrate, then that's what your going to get in the second pass.
Acaila
19th March 2002, 09:29
If divx5 is like other two pass methods, then it doesn't matter what the second pass bitrate is. If you set the bitrate to 1400 on the first pass, the second pass can't be raised higher or lower, it'll just be 1400 as well. The second pass is setting the bitrate from the divx log file. If it's following a log file for 1400 bitrate, then that's what your going to get in the second pass.
What other 2-pass methods are you talking about? Both DivX3 and XviD use a constant quantizer of 2 for the first pass.
DivX4 doesn't stick to this maximum you're talking about either. Lower the bitrate in the second pass and you'll get a smaller file. DivX5 works like DivX4 in this respect. All the first pass does is perform motion estimation on the movie, and although that estimation will be off when you change the bitrate you can still get a bigger or smaller file.
Excelsior
20th March 2002, 22:49
Sorry I expressed myself a little unclear. I meant that it didn't matter what bitrate above 1300 in 2-pass first pass the size didn't get bigger. Thanks for all your answers I got it now by putting a higher resolution and audio bitrate higher I think it was 192 or more I'm not sure. The quality is superb! :)
karloz1
20th March 2002, 23:51
Originally posted by Acaila
What other 2-pass methods are you talking about? Both DivX3 and XviD use a constant quantizer of 2 for the first pass.
DivX4 doesn't stick to this maximum you're talking about either. Lower the bitrate in the second pass and you'll get a smaller file. DivX5 works like DivX4 in this respect. All the first pass does is perform motion estimation on the movie, and although that estimation will be off when you change the bitrate you can still get a bigger or smaller file.
Well, I've never messed with quantizer settings, so I think you're right. But whenever I used AVI revolution 2 pass, which used another program (forgot what the program was called), it didn't matter what I set the second pass bitrate to it wouldn't raise the size of the file size more than a couple mb, or lower it more that a couple mb. So, that was divx3, Cacaila. But, wtf do I know? I was just trying to help this guy out, because no one's given him a response that's helped any (including yours).
dragoman
21st March 2002, 01:54
Actually, I don't think the codec actually encodes all frames with quant 2, it encodes every frame at the highest quant it thinks is the best for that frame.
What? So you are saying that despite the fact that whatever bitrate he sets, he gets a max filesize of 1gb, and you don't "think" the codec is using a quant of 2 to encode every frame?
Bullsh*t. If he does a 1-pass quality based at 100%, he will arrive at the same filesize, and that DOES use a quant of 2 for every frame.
Yes, of course a black frame doesn't need as high a quant as a high action frame, but if the codec is trying to put the most bits is can, and every other frame in the movie is a quant of 2, it's going to encode the black frame at 2.
and karloz, what is this saying I was just trying to help this guy out, because no one's given him a response that's helped any (including yours).
I think I helped him, we all did, which is proved by him saying Thanks for all your answers I got it now
Sounds like we helped him to me...
dragoman
Franky
21st March 2002, 16:01
>Yes, of course a black frame doesn't need as high a quant as a high
>action frame, but if the codec is trying to put the most bits is can,
>and every other frame in the movie is a quant of 2, it's going to
>encode the black frame at 2.
Um, no. That's what's changed in DivX 5, it doesn't use a higher quant than it feels is neccesary for any particular frame regardless of the bitrate, that's why people get undersized files. It will use a lower quant if the bitrate is low though.
tripnotik
22nd March 2002, 00:24
@Franky
I think what you say is true with psychovisual modeling turned on only.
Franky
22nd March 2002, 02:32
Nope, I don't think so, because this is the behavior I see from DivX 5 (or what I believe is DivX 5's behavior from the results I get using it) and I have yet to touch psychovisual. It could very well be one of the other new MPEG-4 settings though.
jonny
22nd March 2002, 10:10
I agree with Franky, i've encoded a movie using a compress test for divx5... the result was 95,9% of compressibility test
When i've done the 2 passes: the video is 58MB undersized & avg quantizer value is 2.89 (with 95,9% should be something between 2-2.5!!!)
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.