Need Help finding matrices

[dark.soft]

Hi, I need to encode 2 different types of video:
-Cartoon with a size between 320x240 and 640x480 (I know that 90% of the cartoon resolutions used are between QVGA and VGA, but I wanted to be specific)
-Movie with a resolution between 640x272 and 640x360 (sometimes 640x480), with a bitrate between 1300 and 2000. So, I need a matrix that can preserve a lot of details. I tried a Matrix 3 encode with these settings (XviD 1.0.3):
bitrate= 1971
bframes 1/1.1/1
VHQ 1 (I read that level 4 doesn't offer more detail)
chroma motion
minQ1n maxQ31 for I\P\B frames
Trellis
MPEG Matrix
Overflow Threatment= 10\5\5
Curve Compression= 10\10

and I didn't obtained the level of detail that I was expecting. For example, when there was dust, it was a bit washed, the face details weren't all there, etc. I compared the resulted file and the original with virtualdub.

So my needs are: one or more matrices for the 2 video types and maybe some hints for the xvid settings that I should use for Matrix 3.

Hope that someone can help me and Thanks in advance :)

[Sharktooth]

for the 1300-2000kb case you can practically use any high-bitrate matrix you can find (except some "extreme" matrices).
You can start with 6of9, 6of9 HVS, EQM V2, EQM V3 HR, EQM V3UHR, HVS Better, HVS Best etc... (use the forum search function to find them).
Remember using a high bitrate matrix can compromise the movie playback on standalone players.
I suppose "Matrix 3" is "Matrix Revolutions"... it's a pain in the ass. I encoded it with x264 coz xvid needed too much bitrate (hence a much bigger filesize) to obtain pleasing results. However, if you want a decent quality, forget about encoding it with xvid in less than 1GB (if you want to keep a res of 720 or 640px).

[dark.soft]

I suppose "Matrix 3" is "Matrix Revolutions"... it's a pain in the ass
You are right: the codec preserves all the bitrate for the final scene with lots of rain, and B frames doesn't help preserving details in that scene.

I encoded it with x264 coz xvid needed too much bitrate (hence a much bigger filesize) to obtain pleasing results.
You're right again! I also tried x264 at the same bitrate of my previous test with XviD (1971), but at full anamorphic DVD resolution (with black bars), and I was blown away! The only thing I didn't like was that the codec smoothed the grain on monochrome backgrounds. But I can't encode in x264 because obviously I can play it only on a computer.

However, if you want a decent quality, forget about encoding it with xvid in less than 1GB
In fact the size that I want is around 2 GB, but at that bitrate I want almost good face, rain and debris detail! Since I can't try to encode with many matrices with my not-so-fast PC, can you tell me what is considered by the community one of the best? Or at least 3 of the best! Thanks.

[Sharktooth]

Try with the ones i posted above.

[yaz]

@dark.soft
if u want full hw compliancy forget about cqm. most hw players do not support that.

your settings is suboptimal (imho) :
- use xvid 1.1
- don't use q1 unless u got heavy undersizing (i bet u'd never hit that w/matrix 3 ;) )
- do use vhq4, it helps. not necessarily on preserving some specific details but in overall.

anyway, if u lost only that (very) fine details u described i'd vote for a 'too high' compression. what is your compression (the rate of 1st pass and 2nd pass size) ?

if u want that fine details don't bother w/the cqms above. they will produce the same or (much) worse. keep up w/mpeg but give enough bitrate. it will be pretty high for matrix 3 but ... sg for sg ;)

once i encoded the matrix trilogy and animatrix onto 1dvd for a friend of mine. i used identical compression for each and, iirc, matrix 3 took up more than the half of the space. after all i must have downsized to keep reasonable compression.

the bests
y

[dark.soft]

Well, the compressibility test returned that I had 94% of the quality of a constant Q2 encode. I can't explain myself why I'm not obtaining more detail.

[*.mp4 guy]

Q2 is far from perfect quality with most matrices, but it is the highest reasonable quality you can reach with Xvid, or most other Mpeg4 ASP codecs. You will loose some grain no matter what because it is mostly noise and codecs try to give bitrate to the other parts of the picture first, fine details get mistaken for noise sometimes so you will loose a little bit of quality with every lossy encode of a source. IMO the 3 best matrices for your situation most likely are 6of9, EQM V3 HR, and HVS Best.

[yaz]

Q2 is far from perfect quality with most matrices ...ehrm ... and with which matrix do u think it's closer (or the closest) to the 'perfect quality' ?
... IMO the 3 best matrices for your situation most likely are 6of9, EQM V3 HR, and HVS Best.in my very last test (w/a different source) these matrices produced 1st pass sizes roughly as 4.5 / 2.5 / 1, respectively. how do u think they could be interchangable in this situation ? what u may gain on the 'more detail preserving power' of a matrix (say, 6o9) u would loose on the (much) higher compression u should apply.

anyway, loosing fine details may have so many reasons, and lossy encoding is only one of them. u should also consider : (heavy) filtering, (lossy) resizing (on encoding as well as on playback !), color space conversion(s), aso, aso.
the compressibility test returned that I had 94% of the quality of a constant Q2 encodewhat does this mean in english ? u want to do a 94% compression ? or what exactly ?

the bests
y

[dark.soft]

what does this mean in english ? u want to do a 94% compression ? or what exactly ? Sorry, I know that I don't speak a perfect English. What I wanted to say is that I had a compression of 94% with a bitrate of 1971 (this is what gordianknot said after a compressibility test of 5% of the movie).

in my very last test (w/a different source) these matrices produced 1st pass sizes roughly as 4.5 / 2.5 / 1, respectively. how do u think they could be interchangable in this situation ? what u may gain on the 'more detail preserving power' of a matrix (say, 6o9) u would loose on the (much) higher compression u should apply. So, what matrix should I use with the bitrate I'm aiming for?

[*.mp4 guy]

ehrm ... and with which matrix do u think it's closer (or the closest) to the 'perfect quality' ?
in my very last test (w/a different source) these matrices produced 1st pass sizes roughly as 4.5 / 2.5 / 1, respectively. how do u think they could be interchangable in this situation ?

All I was saying is you will _Always_ loose details when using xvid, especialy with a matrix like Mpeg or H.263. the matrices are interchangeable in this situation because because they all preserve good details but in different situations, he needs to find which one he thinks looks best on this encode, its all subjective.

[feedback]

Well you asked so Here is a link to Matrix zip folders. (http://forum.doom9.org/showpost.php?p=685833&postcount=23)

Regards,:)

[dark.soft]

I've tried EQM V3 HR with a telefilm encode because I hadn't got the time today to do a full feature encode. I must say that I'm satisfied, it preserved a good amount of detail, so I'll try more V3 matrices.