FINAL FANTASY ... impossible movie??

[NoFace319]

Hi guys... I have a big problem with Fimal Fantasy (the Film movie)...
The F*cking movie has too many colour variations... so... I think that it's impossible 2 encode it in a decent quality and without that F*cking squares.... I'm about to go mad guys... I can't encode it in a decent way... here's my Setting in Goardina Knot:

Resolution : 704 x 368
I did the compressibility test : 69% of 0,325 that means a 0,227 bits/(pixel*frame) <-- Correct?? ok... It fits on a 2CD's.


I did it with DIVx 5 and DIV3... and the black and grey and... aaaaaaaaaarg U know what I mean... too much shit... the shadows.... all is a shit.... and guys I have the same problem with "Gladiator"... andother crazy movie.... look at the first sequence... Crazy man...

Well... I really need help... I'm following all the steps and reading all I can do... but... I can't with those movies... I can't think about encoding "THE OTHERS"!!!!!!!!!!! All is mostly black...

Anybody can give me a solution??? or a settings or a paht 2 follow!???


PD: I used VIrtualDub at 1 Quality pass ... (Quant 2) and continues that shit...


Thanks in advance.... Kisses from Spanish Female

[NoFace319]

and I forgot 2 post the bitrate....


1473kbit/s


Thanks

[Acaila]

First of all, reduce the resolution by a heap! If you can't cut it at 704 reduce it to 640. I've done Gladiator at 640 on 2 CD's (with AC3 audio) and it came out beautifull (although it was XviD, not DivX5 :)) and I'm sure others have done it as well.

Although 69% should be enough for a normal movie, FF is a computer animation and thus like cartoons needs more bitrate to look good so this number doesn't really say a lot for this movie.

You could use a TemporalSmoother(1) on FF, since it's animation you probably won't notice it anyway. Don't do this on Gladiator or you'll kill that superb movie.

I don't know what audio you are using, but here's a few suggestions:
- If you plan on using AC3, then if the bitrate is 448kbps you could re-encode that (with AC3-Machine) into 384kps to save some space.
- Or you could encode the AC3 into Ogg Vorbis if you want to save a lot of space.

Ps.
All DivX codecs suffer from a macroblock problem in shadows. There is nothing you can do about this, if this is what you are referring to. On TV the effect is much worse than on a computer monitor though.

[DJ Bobo]

you're a funny one, encoded with quality based 100% and not satisfied, what would satisfy you?!
I think your post processing level is too low, that's all, push it to max or near max and you'll be satisfied I guess/hope -.-

[NoFace319]

Now testing "Gladiator" with Xvid following your rules... let's see the results...
I'll tell ya in a while.

Thanks ;) :rolleyes:

[grug2k]

I don't recall having a problem with Final Fantasy. I think I got it on 1 CD with ~80kbps OGG and 640 or 608 res.

[theReal]

From what you say (69% compress test, then 100% quality, but the results are still looking awful) I think you're having a playback problem! Maybe some kind of overlay problem with you graphics adapter. Try to keep the video width and height divisible by 16, try other players, other graphics-card drivers and set your desktop color depth to 32 or 24 bit.

[NoFace319]

I know that this reply soudn't fint here but as u recomend me Xvid 2 encode it I have decided 2 poste here continuing my first question.

Gladiator: Encoded at 640... 2 cd's... XVID... AC3 medum quality 384 Using at the first pass H.263 and Modulated at the seccond pass... Now I think that everything is all right (any suggestion about that???) BUT shall we talk about ALT curves

Do u remember your own ALT CURVES used in Gladiator... LOT OF Movement at the begining of the movie (the battle)... and I put MEDIUM AGRESSION, 200 HIGH, 100 Low.. 50% streng...and looks... awful btw I have read a lot of alt curves in here but I don't know what parameters I have 2 put... I'm a little confused (my english isn't quite great alt all hehehe) Do u remember your own parameters for Gladiator???

PD: My GRAPHIC CARD is... GEFORCE 4 4400 128mb... No problems I think... The People that have encoded Final Fantasy please take a look at the first 5 minutes specially at the DREAMS MACHINE... when she floats on the room... Absolutely AWFUL in my test encodes... if u please give me your own parameters (bitrate, Pixel Rate, Resolution.. etc.. I WOUL APRECIATE A LOOOOOOOT)

Thank U guys... every day I learn a lot with you...

KISSES FROM SPAIN!!!

:)

[Acaila]

I know that this reply soudn't fint here but as u recomend me Xvid 2 encode it I have decided 2 poste here continuing my first question.Actually I didn't advise you to use XviD, I was just telling you what -I- used to encode this movie :).
And I'm sorry, but I don't remember which CC settings I used (I know I didn't use Alt-CC, frankly I don't like it). Nothing wrong with your settings so far (although MPEG on 1st pass would probably be more efficient than H.263, depending on how many frames get quants of 2-3), might wanna try Lumi Masking though.

For further discussion on Alt-CC settings, please post in the XviD forum (or PM me ;)).

[NoFace319]

So what's the best encoder?? I like Xvid a lot... The macroblocks stay in its position instead of keep moving every time like happens in Divx5/4/3... Anyway thanks 4 your advices :) I make another gladiator movie... MPEG first pass... Modulated seccond pass... and... well... the macroblocks in the action scenes.. still there

:_( hehehe... keep tryng

Kisses

[guirlle]

just to say that I did this movie whith Fairuse just to prove (long time ago) and the results were spectaculars, compared whith other movies (no as spectaculars as the actual DivX5 movies ...). So I think that have to be possible to do a good movie with DivX5. Becouse i have this DVD I'll make a test this weekend...I put the movie in my job list.

BYE

[NoFace319]

Thanks man :)

Gracias corazón :):cool:

[Darksoul71]

@NoFace319:
You´re really funny ! 100% quality isn´t good enough for you ?
Well, then buy the DVD.

OK, DivX is a lossy format. So you have to live with some loss in quality. If you can´t live with this quality loss then buy the DVD.

As Acaila already mentioned: Lower your resolution to somewhat around 640 pixels width. Also don´t use a too high bicubic filter. DivX doesn´t like this as sharp corners & edges need more bitrate.

Try the following:
1) Resolution to 640x.... Soft or Medium Bicubic (if you use GKnot for Resizing)
2) Add a TemporalSmoother (1) or even (2) to the AVS generated by GKnot between crop and resize
3) Encode using DivX5Pro with GMC and bidirectional encoding enabled

One CD should be "good enough" unless you have VERY HIGH quality requirements. I did FF a while ago with DivX 4.12 (as far as I could remember) in 2 Pass as 1 CD movie and was very pleased with the results.

MOST IMPORTANT: Use a medium to high post filter during playback because DivX4/5 really needs this. Only while playing back on TV you can forget about postfiltering because the TV screen softens the image enough.

So what's the best encoder??
That is up to your personal viewing behaviours. Here from my point of view:

XVid/DivX 3.11 (with Nandub): Overall sharper image which means that you can use a lower res when compared to DivX. More tendency to build macroblocks (depending on the encoding settings you use)
DivX4/5/etc: Softer image (also depending on postprocessing settings). Higher res / bicubic filters needed when compared to XVid & DivX 3.11. Better able to work with noisy sources. Good postfilter (blocks out a lot of artefacts).

In general it depends on: Movie, bitrate, resolution, filters, etc.

So you see: There isn´t the "best" encoder ! Unfortunately but IMHO DivX provides better results for newbies because XVid and DivX 3.11 (Nandub) need a lot of knowledge to give good results.

I like Xvid a lot... The macroblocks stay in its position instead of keep moving every time like happens in Divx5/4/3

Hm, you like Xvid a lot because "the macroblocks stay in its position instead of keep moving around" ?

I hope you know where the macroblocks come from ? It´s because the encoder (not matter if it´s DivX/XVid/MPEG1/etc) isn´t able to compress the block (macroblock) correctly. Be it because there´s too much movement or too low bitrate. And think of an average movie: Does it make sense that "macroblocks stay in position" ? No, they will move like the area of movement changes with every frame.

@Acaila: Just a few comments:
You could use a TemporalSmoother(1) on FF, since it's animation you probably won't notice it anyway. Don't do this on Gladiator or you'll kill that superb movie.

Hm, I haven´t done Gladiator by using Temporal Smoother but a Temporal Smoother at Level 1 or 2 between cropping and resize isn´t really noticable. It helps to make the movie more compressable and I often use this on tricky movies (much movements, long duration for 1 CD, etc). If you use a TempSmoother after resize it will generally be more visible.
DivX smoothes out a lot of details anyway. So another "few" details removed isn´t a big deal here. As I said above: Those who want max quality should buy the DVD :)

All DivX codecs suffer from a macroblock problem in shadows. There is nothing you can do about this, if this is what you are referring to. On TV the effect is much worse than on a computer monitor though.

Those macroblock problems are both good visible on computer monitor AND TV equally. You can mainly see them good if you bumped up your brigthness too much. Lowering brightness can do wonders.

Just my two cents,
D$

[gldblade]

>Does it make sense that "macroblocks stay in position" ? No, they will move like the area of movement changes with every frame.

Yes it makes sense if you're talking about things such as walls. Or even someone's face when they're staying still, such as during a conversation. It wouldn't make sense to have macroblocks move in stationary objects, would it?

[Darksoul71]

Of course, in this case it makes sense but on a wall (esp. a quite equal colored wall) macroblocks tend to move around with XVid also.

So unless you have a VERY STATIC scene (a face not moving in the same camera perspective or a wall with no moving objects and also the same camera perspective) macroblocks tend to float around, wouldn´t you agree ?

And: In a well encoded movie macroblocks will only show in maximum stress situations for the codec.
So basicly something like smoke, fog and other quite equaly colored parts of the movie.

-D$

[Acaila]

@Darksoul71:
Hm, I haven´t done Gladiator by using Temporal Smoother but a Temporal Smoother at Level 1 or 2 between cropping and resize isn´t really noticable. It helps to make the movie more compressable and I often use this on tricky movies (much movements, long duration for 1 CD, etc). If you use a TempSmoother after resize it will generally be more visible.I agree with you, a TempSmooth(1) will probably not be visible if you bump post-processing up to halfway like you said (I always watch without post-processing). I also use if for movies that are a tight fit for 1 CD, but generally I try not to use it. For a movie like Gladiator that fits perfectly on 2 CD's there's really no need for it. Especially not for XviD like I did.
And: In a well encoded movie macroblocks will only show in maximum stress situations for the codec.
So basicly something like smoke, fog and other quite equaly colored parts of the movie.Yeah, I hate smoke :D

The only codec I have found to be very good against the "moving wall" phenomenon is DivX3. Others like DivX5 and XviD all seemed to have it, but DivX3 is very good at keeping things static. DivX4 on the other hand blurs way too much, so I call that cheating :).

[DJ Bobo]

I can't say it enough times: don't use GMC cause it's buggy. Somewhere someone prooved it in this forum.

[theReal]

It helps to make the movie more compressable and I often use this on tricky movies (much movements, long duration for 1 CD, etc). Doesn't really belong here, but I just noticed that there are some movies where this doesn't help. I tried temporalsmoother settings (1), (2) and (2,1) and I even tried light and normal pre-processing in Divx5 but all I achieved was a compress test that went from 38% to 39%.
Maybe that was due to the fact that I was already using VDub's precise bilinear filter for resizing, maybe it was because the movie doesn't have analog noise, but mpeg2 compression artefacts...
I'm talking about the Big Lebowski (PAL, 16:9) - the DVD quality is really bad, some scenes almost look like a Divx rip with mosquitos and macroblocks.
I really don't want to waste 2 CDs on that movie, but I wanted to have two audio tracks. Well, maybe I just settle for subtitles instead of 2nd audio :)

[theReal]

I can't say it enough times: don't use GMC cause it's buggy. Somewhere someone prooved it in this forum.I think that was in Divx 5.00 and maybe still in 5.01. I experienced some problems myself in 5.00, but I've always used gmc since 5.02 and never had any bad results.

[unplugged]

Final Fantasy is a very noisy movie (artifical noise of course :)), I have performed it with XviD, particularly convinced to use H.263 quant. matrix (noise tend to generate macroblock color fluctuations).

XviD 28-06-2002

H.263 quant. (to greatly elminate partly the noise ;))
Max I-Frame interval: 1000 (40 seconds, no problems with OGG container ;) monster seeking)
Min I-Frame interval: 1
Min P-Frame quant: 2
Max P-Frame quant: 31
Alt. Medium curve (linear is more predictable and less weird, guys!)
High curve: 350
Low curve: 85
Curve auto strength 0% (no curve compression over my trusted settings 350/85)
both MV Hints and lumafix DISABLED

Resolution: 704x544 (often little higher quantizers at high resolution keep mooore detail than low at lower res, plus I hate resize with its N implications)

1st pass (2 quant) test: 2,450,000 Kbytes
2nd pass target: 1,372,000 Kbytes
(+ Vorbis audio track VBR ~90 Kbit around 60 Mbytes = ~ 1400 Mbytes)

This is a very compressible movie considering ~1,5 hours with the very limited crop zone 704x544, because of course the 1st-pass size has reached "only" 2,450,000 Kbytes.

The result compared frame by frame with the original (Vdub .avs mpeg2dec script at 704x544) is very detailed as crisp is the source!
Stable and linear quantizer fluctuation and must admit that quant 3-4 is outstanding at 704x544.
Playback perfect even with Celeron 800 CPUs!
(plus I don't use postfilter even with my K7 2100+, else it will clean out 30% precius detail)

My little "transcoding" tricks:

- Crop OFFSET (x,y) aligned to 8x8 pixels to match MPEG-2 macroblocks, thus maintaining crisp detail inside macroblocks (you don't imagine how this "stupid" thing has its own specular result) without blurring sharp edges between these 8x8 cells.
- Crop SIZE multiple by 32 pixels for X (704) and at *least* 16 for Y (544) to optimize internal CODEC efficiency
- Direct YV12 MPEG2-decoding -> MPEG4-encoding thru last MPEG2AVI revised by Iker

[NeVeRLiFt]

After you ripped enough DVD's you will learn some movies just compress and look better than others.

I can make any movie thats 2hrs or less long a 1cd rip and have it look good.
Take a look at my settings and try it. I'm sure if I can make a high quality 1cd rip and use 640x288 or 512x384 then you should beable to make a good 2cd rip ;)

http://forum.doom9.org/showthread.php?s=&threadid=29134

http://forum.doom9.org/showthread.php?s=&threadid=29221

http://forum.doom9.org/showthread.php?s=&threadid=29206

[Dagon]

Are you kidding? So far Final Fantasy is one of the best looking movie's I've done. Divx 3.11a and gknot and I can't the diff between the dvd and the rip, and I'm a picky bastard when it comes to visual quality.

[unplugged]

I have only said that the original MPEG2 content has *artificial* background noise (film effect, understand?), not relevant BUT maybe a pain for the encoder.
Partially solved using H.263 quantizer instead of MPEG.

Yes, the overall DVD quality is stunning.

[vorbis]

lower the resolution !!
I remember doing that one in divx 3 at a fairly low resolution 544 for a 1 cd rip. It looks pretty good, friends can't tell its not the dvd. Maybe you simply have too high standards for divx. Also am I the only one who thinks you can get good rips without resolutions of 640 and upwards.

[TheXung]

About temporal smoothing:

Temporal smoothing isn't some technique that you can apply blindly to any movie. Infact, temporal smoothing if done blindly usually results in less quality. In any scene where there's a decent amount of movement, when I say decent, I'm not talking alot, I'm mean just any area of the screen where there's something moving, temporal smoothing will result in more noise in the picture. At a fixed quantizer, it may reduce filesize by 10% but even if you lowered the quantizer to use up that 10%, the image still has more noise in moving parts. Freeze the frames and see for yourself. The noise is even worse in animated and computer graphics. Temporal smoothing needs to be applied adaptively and pretty much inside the codec to improve quality.

[NoFace319]

Nice man... I'll try ... :) hard parameters... :)

[kilg0r3]

@unplugged

i know this comes a little late but, how exactly does this macroblock alignment thing work? Does it simply mean that my cropping values must be divisible by 8? And that i am not allowed to resize at all? thaks for your patience in advance.

[unplugged]

Must be said that macroblock alignment is to be considered as optimization *only* if the image remain the same, NOT resized (sure) or anyway filtered/converted (IMHO).

Simply every MPEG/JPEG content has the compression scheme arranged by internal 8x8 *windows* or cells, these can be seen (macroblocks) mostly when the content is highly compressed and/or zoomed.
Doing some quick view myself I have seen that common MPEG2 content has visible compression artifacts as these macroblocks doesn't joint very well together especially when compression show itself :).
As these blocks can create a pretty sensible "chessboard" artifact not only for human's eye but also for the "encoder's eye", when the content is reencoded with XxY crop points NOT 8x8 aligned the compression scheme tend to sample too much false detail speculated from these sharp-contrast artifacts since each new re-encoded macroblock is based on area across 4 macrobloks, this wastes bit allocation and worsen the output image detail.

I have made some parallel tests at fixed quantizer 2 and 3 with XviD, focusing to this X,Y crop offset issue (and using X res. multiple by 32 and Y almost by 16... but this is another argument, already discussed):

50% of files re-encoded with 8x8 block matching/alignment are 1-2% bigger than the others aligned 4x4 (ouch... each macroblock perfectly across the 4 around).
The other 50% even 2% shorter.

All the files encoded with 8x8 alignment looks better detailed than the others and especially very to close to the source.

Conclusion: there is little more detail, but better said: the real detail.


P.S.: Plus, I'm not sure... but if macroblock, due its native quantization mechanism, tend to create "false" adaptive pattern(s), block matching/alignment is also a good thing to serve clearly "this already distorted pattern" to the MPEG re-encoder, almost to prevent yet another (adaptive) distorsion.

[unplugged]

Originally posted by kilg0r3
Does it simply mean that my cropping values must be divisible by 8?

Finally yes, your (Left,Top) X,Y cropping values.
For resolution it's advisable to adopt X divisible by 32, Y by 16. (or better... 32)

[Jas113]

This movie was my first rip. I did it in FLaskMPEG w/ divx 4.12 using one pass bitraet 1800 (2 cds). Its my best looking movie. No filters, no BS. Its it easily DVD quality.