an old codec beats a new codec?

[dimitrius]

I have this old (5 year or so) encode of Leon (uncut), done in divx3. The movie is about 2 hours long, can't remember the res of the top of my head (about 480 to 512 horiz.) and there are action scenes, explosions, fast moving bodies (from tops of buildings to the bottoms). The whole thing fits into 650megs and the quality is pretty good considering the size and the length.

Now, I thought, I'd rent the dvd from blockbuster and rip it with something alittle more mordern into 700megs of near perfect quality. But, no matter what I do I just can't make it look right (especially the big explosion at the end). I have basically resorted to using genetic algorythms to get the right settings (of course I made sure to reread xvid and divx guides). Still, it never looks quite right.

Is it me or is that modern builds of xvid and divx haven't really moved on that much. 2-3 years ago it seemed like by 2005 we would fit a dvd on a floppy but it's going slower than i thought it would.

what are your views on this.

[Tuning]

Is it me or is that modern builds of xvid and divx haven't really moved on that much. 2-3 years ago it seemed like by 2005 we would fit a dvd on a floppy but it's going slower than i thought it would.
:D
I don't think developement going backwards.Everything is moving forward as the case of XviD/DivX.The new builds may be slower in encoding but they are defenitely better than old ones.

I think the problem is with the encoding method you are employing.Can you explain a little bit more how are you doing this...

Tuning

[dimitrius]

well, i am not saying it's going backwards, just slower than expected, i think.

as for that movie, i've done 7 different encodes with xvid, and 3 with divx 5.1 (dam it the slowest setting really means very slow...). always used 6 - Ultra High, tried both quantization types, B frames with setting of 2, 3 and none (i think that was -1) and tried VHQ. I am not saying the result was bad, but worse than the original (you see the original had some grain effect which made it look offentic like those remastered western dvds - so that didn't help i guess), and certainly I was kind of expecting more after 5 years.

Don't take my post too seriously, I am just babling

[Tuning]

Well pardon me .... you being senior than me on using DivX..

It is true that DivX is getting slower and slower as version ascends.Of them DiVx 5.1 is slowest of all.But there are major quality improvements and the new build is using special algorithms.

And don't think we are not using DivX 3.11 now.I'm a big fan of DiV3 as I could encode 2hr movie in less than 5hr on my P4 1.4ghz machine in two pass.

but worse than the original
Are you reffering the original DVD,Or your first rip?
Then I think the source is different than what you used earlier and that is causing diminished details..(May be)

Just for curiosity may I ask :Is there was Avisynth at your time?
If not how do you done your encodes.

[dimitrius]

well "but worse than the original" I mean the dvd i rented from blockbuster. Also that first 650meg encoded wasn't mine, someone uploaded it to my server when i used run one on dalnet on irc.

I've never actually encoded anything with divx3 simply because I consider it old - maybe I am wrong about that in which case it would go well with the topic of the post.

you are quite right about divx5.1 - really slow (i have an athlon xp 3000+ - about 2167ghz and ddr333 and encoding to a raid array and still I can only get 4-8 fps)

avisynth? don't think anyone used any frame servers then...

[gurabli]

I think too that the problem is the source - DVD.
I have seen the same movie on a different DVD and it was not the same.

It is just impossible that DivX 5.0.X or DivX 5.1 gives worse quality than divx 3X.
Maybe not in the case of Divx5.1 because it needs higher bitrates to produce extreamly good quality - the best ever.
In case of XviD - it must be the setting you used to encode.
Try 1 or 2 b-frames, VHQ 4 (very slow - but very good quality) and dont use lum.mask.

Something else,
the problem can be the size of your audio file!
What is the bitrate of your divx3 file audio?
If you encoded with higher bitrate (or maybe ac3 what I do not think, because it is almost impossible to do with 1CD rip) it can be the problem.
Check the audio bitrate of your divx3 copy and then compare with the one you made now.
Just guessing, and giving a few tips.

[symonjfox]

I remember I used DVD2AVI for Demux the audio and create D2V file. Then I use VFAPI converter and Virtual Dub for encode to avi.

As far I can say, there have been made TONS of improvements since Div3. For example the High or Low motion was a kind of matter, I always had to encode it twice to see which was the best.

DivX 5 and Xvid are far better than DivX3, you just have to tweak it a little bit to get better results (Xvid).

And DivX and Xvid now are better MPEG4 compatible, while DivX3 is NOT!

PS: I won't complain about speed. Speed is relative. Today it's slow, tomorrow it will be fast, but there will be another kind of slower compression and so on ...

[Mango Madness]

software tends to get slower, it just happens. the tradeoff is for better quality products. The software industry assumes that computers get faster, which is a pretty good assumption. Once advanced simple profile is exhausted, we'll probably move on to h.264 which will add another level of slowness but hopefully by that time we'll have a new generation of hardware. The key things to pay attention to is with compiler technology and new, faster mathematical algorithms. Just don't expect huge increases in speed because the long-term trend is towards slowing down with the result of higher quality encodes. If software didn't push hardware, then AMD and Intel would be out of a job. These are extreme generalizations, so take them as such.

[seewen]

Originally posted by dimitrius
I am not saying the result was bad, but worse than the original

Are you ABSOLUTELY CERTAIN that :
- You are using the exact same resolution.
- In case you are using Avisynth, are you using the exact same script (same resizer, same filters, etc.. ).

If you don't, you cannot compare your old DivX 3 with your new XviD/DivX 5.

[ChristianHJW]

Some of the DivX3 movies you find on p2p are so good, you have to assume that some professional digital video equipment was used to prefilter the source material ;) .... and professional filtering can do wonders ....

[Koepi]

Professional filters can't do wonders. look at all the crappy DVDs out there. Oversharped, too smooth, artefacts from temporal smoothing which we don't produce as we're having filters like C3D,... and so on.

There must be soe other magic involved in such DivX movies :)

Regards
Koepi

[Prettz]

If a divx3 encode looks really really nice it was most likely encoded with Nandub by someone with a lot of experience with using it, and they probably scanned the whole movie with a fine comb and forced any questionable-looking frames to keyframes. If you spend that much time doing it you can probably get the same result with xvid. Of course your result will still have an "xvid-look" to it rather than an "SBC-look".

[Rash]

Originally posted by Koepi
Professional filters can't do wonders. look at all the crappy DVDs out there. Oversharped, too smooth, artefacts from temporal smoothing which we don't produce as we're having filters like C3D,... and so on.

There must be soe other magic involved in such DivX movies :)
I agree Koepi. Some DVDs have horrible image quality. I can tell you it is worse than VHS or old mpeg-1.

And my Back to The Future encodes came out perfect! :D The sources were horrible, but I used all the filters I had and suggestions from people here in the forum and it came out really nice. So maybe you should try some more things Dimitrius.

[dragongodz]

sorry guys but you do realise this thread should have been closed already dont you ?

the divx 3 copy is one uploaded and the divx 5 one dimitrius has tried to do is a rented dvd. so neither can be claimed to be for backup etc.

[dimitrius]

sorry guys but you do realise this thread should have been closed already dont you ?

turn on cover fire...
eh, the div3 copy was downloaded before RIAA really cared about any of this. I tend to become abit obssesive about try to achive something and at that time I was compering different codecs (renting was a cheaper solution to do the tests). Recently I have bought the actual dvd of Leon spent the whole £16 on it (I do actually make it a point to buy movies I like - Leon, LOTR, Star Trek Nemesis etc.). Sorry I haven't posted much on this board so wasn't used to the whole "backup only, no pirating" thingy - I'll be more dilligent about what I say from now.
turn off cover fire...

as for the practical side of the thread - I was just encoding the video at that point which I think was about 580 megs so if we ignore overheads the audio is about 70 megs. yes it was the same resolution. it is pretty easy to get excellent quality at higher bitrates but for this size (the movie is about 2 hours long so bitrate would be about 660 kbits/sec) I have not been able to beat the div3 encode in terms of quality (I havent tried wmv9). i guess it was encoded by a real expert and there weren't any kind of comments in the avi.

as for theoretical side...
I think that alot of stuff that came out recently was mainly improvements on the old stuff and very little in terms of new stuff. divx 5.1 just brought in a nicely working psych. enh. engine. xvid kind of made luma masking work more or less ok. however there hasn't been anything profoundly new (though ogg container is now more popular).

I am actually working on an idea that in theory might give almost lossless compression at CD sizes. It involves genetic algorythms on the encoding side and 2 neural networks on decoding. though by the time i make anything practical out of this we'll probably have holographic storage in which case noone will care about the size. The idea is also to make the system on hardware. so you would have this pci card that the encoder and the player can use. The card would have genetic algorythm processor, and some analog electronics for the neural networks. But that's just an idea for now.

Interesting that noone took up fractal compression anywhere and it was a buzzing topic for quite sometime at several UK universities.

[Tuning]

I am actually working on an idea that in theory might give almost lossless compression at CD sizes. It involves genetic algorythms on the encoding side and 2 neural networks on decoding. though by the time i make anything practical out of this we'll probably have holographic storage in which case noone will care about the size. The idea is also to make the system on hardware. so you would have this pci card that the encoder and the player can use. The card would have genetic algorythm processor, and some analog electronics for the neural networks. But that's just an idea for now.

Interesting.....

by the time i make anything practical out of this we'll probably have holographic storage in which case noone will care about the size.

streaming will?.Future proof?

[dimitrius]

streaming...
yeah that's the subproblem i am thinking about now. With the system i have in mind now you can't stream and you can't even cut the movie without almost total reencode. but on the plus side because of the way genetic algorythms work the longer you encode, the less data you loose from video, the better the result.

maybe I'll make this into my PhD...

I'll be Doctor Gofman perhaps :cool:

[Tuning]

Wow!
You are far more than I thought.Hope your Research carries on and get your PhD.

Best wishes Gofman!.

-Tuning

[Mango Madness]

we've brought up evolutionary computing along with AI algorithms in the chatroom before and determined that it's beyond the scope of xvid. A university with tighter control over coding and hardware would be better suited for such an application. The problem we theorized, actually MfA brought this up, was that the solution space involved with the pixels in a movie would be too large for conventional computing (ie single user). Eventually though, someone will have to do the ground work for future "super" codecs and I believe this lies with some sort of combination between evolutionary computing and neural network modelling.

[dragongodz]

"the div3 copy was downloaded before RIAA really cared about any of this"
irrelevant. you asked the question here and now.
"Recently I have bought the actual dvd of Leon"
you asked about comparing it to a rented dvd to start with though.
"wasn't used to the whole "backup only, no pirating" thingy"
its pretty well stated in rule 6 so you may want to go brush up on them so you dont accidently break one in the future and get sanctioned.

[birdy]

Originally posted by dragongodz
"the div3 copy was downloaded before RIAA really cared about any of this"
irrelevant. you asked the question here and now.
"Recently I have bought the actual dvd of Leon"
you asked about comparing it to a rented dvd to start with though.
"wasn't used to the whole "backup only, no pirating" thingy"
its pretty well stated in rule 6 so you may want to go brush up on them so you dont accidently break one in the future and get sanctioned.


Oh god! Why the hell do I do this!?
I will have problems and angry people on my back again!
But I guess this is how I am! I can't shut my mouth!

There are always some people, who seem board to death or can't understand anything interesting in a posting then, making there interesting by finding the words, (download, rented etc) in the posting!

Comeon guys,
Its very interesting and honest to see people and to know this board is really against pirating. (I am personaly proud of this way of respecting people who work hard and make softwares and movies and ....)

BUT come on, do you think that sometimes some people overdo it!

I have a question!

If I post a message saying: I downloaded an XVID file and I can't play it back, couldn't it mean, I downloaded a damn amateur movie that some guys did himself and distrubuted for free on the net?

Couldn't it mean that I downloaded a movie from my friend (his birthday) and now I can not play it back?!!!

Damn I KNOW. There are always people who steel movies, software and even if they could they would steel your money or wallet!
But it would be very nice if WE could stop looking at everything with our BAD BAD eye. Lets trust people a little more!

Also is it forbiden, if I want to learn and make experiences with encoding to rent a DVD, rip, encode and then hit del on the final file!?

Or even rent the dvd, rip it on hdd and watch it 5 nights later and then delete?

anyway all this is to say:
EVeryone is not always the bad guy and everything done is not always the bad thing!
Lets just relax a little. We are not the Police, or god or judg or who ever is here to control peoples doings!

Sorry for this!
Ban, insult or say whatever you like now!

[digitalman]

:) Sorry but I have to do this off-topic flame:

" Also is it forbiden, if I want to learn and make experiences with encoding to rent a DVD, rip, encode and then hit del on the final file!?"

Is it okay to want to learn and make experiences with robbing a house, taking their money, and then returning it immediately, just to learn how?

"Or even rent the dvd, rip it on hdd and watch it 5 nights later and then delete?"

Or even steal their money, store it at your house and spend it five nights later and then return it? The DVD companies want you to buy it to watch it later, not make an illegal copy and watch it later. The DVD rental place wants you to rent it for extra days. You are in a way stealing their money.

These may be extreme analogies, but you get the idea. :)

Sorry moderators but delete this flame if so desired.

p.s. - I am just joking around with you birdy. I understand where you are coming from.

[trbarry]

we've brought up evolutionary computing along with AI algorithms in the chatroom before and determined that it's beyond the scope of xvid. A university with tighter control over coding and hardware would be better suited for such an application. The problem we theorized, actually MfA brought this up, was that the solution space involved with the pixels in a movie would be too large for conventional computing (ie single user).

I know very little about neural nets but I've written some genetic algorithm programming code before, back when I was playing with the stock market (and losing). Genetic algorithms are very good for searching large spaces in a non-linear manner. And I think they can scale well to multiple processors or boxes since you can have a population of "individuals" that go off to another box to get processed and evaluated and then report back with a fitness value and the results.

For instance you could do one GOP at a time simultaneously on a few dozen machines. After each GOP you could save the best and use the fitness values to determine what to do on the next GOP. By the time you were done encoding you would likely have a set of encoding paramters that well matched the individual movie you were encoding, maybe even if you did it in one pass and just saved the results.

Sorry if I'm drifting off topic but it was brought up above and it is something I've spent some time thinking about. A lot of good compression techniques seem very close to just trying a bunch of things and keeping the best one.

- Tom

[Nic]

I think I missed something:
"Now, I thought, I'd rent the dvd from blockbuster..."

I.e. if you want a perfect copy of Leon, why not buy the DVD ? :rolleyes:

We don't support anything to do with piracy here, no matter the discussion...

Also, this thread gets subjective, as divx3, although it looks good, looks bad in some areas...and some really don't like that and prefer the way the new DivX/XviD looks. It's all a matter of opinion...Use whats best for you...

Thread Closed.

-Nic

ps
You would have been striked, if I'd have caught this earlier BTW.