IMPORTANT NOTICE;if you toyed with same codecs (as i used in any of these tests) and produced the results that significantly vary from my results,be a sport a start a thread in
Unofficial Virtualdub Support Forums ,so we can discuss it further and so i can update my pages if there's a need (ie. if i was wrong about some codec's performance...)
thanks........

now sit back and enjoy the fruits of my labour.....

mpeg1 at 1,5Mbit/s[CBR] by Tmpgenc
vp6 at 1,5Mbit/s,one pass,best quality
ffvfw image from previous test

previous tests (complementary to this one);
mpeg4_anim-gif.htm
rv9,wmv9,sbc,ffvfw.htm

first of all a small tip;you can use irfanview (i use version 3.10) to split this animated gif into separate images if it would be easier for you to compare some of these shots with some of previous ones that were also integrated into animated gif....[Options->Extract all frames (choose a directory)]

notes;remember how DXN crew claims that divx5 has 10x effectiveness of mpeg2?
i myself assumed that mpeg1/2 might be 2x less effective than mpeg4,but i too was wrong
differences between mpeg1/2 and mpeg4 are not the "2x" figure...it's less than that!!! [i'll be reffering to it as "mpeg1/2" as mpeg2 differs from 1 only in the fact that it does interlaced encoding;performance on progressive clips is the same...my tests prove that,and here's the quote from mpeg official web-site;

"6.1.1 Coding Efficiency Tests
a) Low and Medium Bit rates (version 1)
In this Low and Medium Bitrates Test, frame-based sequences were examined, with MPEG-1 as a reference. (MPEG-2 would be identical for the progressive sequences used, except that MPEG-1 is a bit more efficient as it uses less overhead for header information)." ]

but hold your horses,there's more to video compression than sharpness (in which mpeg1/2 is a clear winner,as illustrated in above gif image......mpeg1/2 wins sharpness test with ease...) :

mpeg1/2 ME algo is usually visibly worse than that of mpeg4,and the lower you go with the bitrate,the more obvious this becomes.....i have also done 448x336 @770kbit/s and compared mpeg1/2 to mpeg4.....no doubt,ME performance of mpeg1/2 deteriorates much more rapidly than mpeg4's.....although spatial performance(sharpness of stiller scenes) stays excellent with mpeg1/2 even on that low bitrates...

in the above image,this is visible as the chroma smearing/blocking around the borders of the image (where the video image and graphics intersect) and also around the "NEU" graphic (to the left of the serial number of the item (194 992) )
mpeg1/2 has more of this artefacts than mpeg4,ie. mpeg4 is more pleasing to watch than mpeg1/2 (although mpeg1/2 still has sharper image,but the effect is diminished with this blocking problems)
so there's no magic;i didn't found my sharpness comfort even with mpeg1/2 (Tmpgenc is also mighty slow encoder) as ME problems become more and more apparent on lower bitrates....
offcourse the higher you go with the bitrate,the more visually pleasing mpeg1/2 becomes....

now,let's see the vp6;test on 1,5Mbit/s obviously show this is a new lo-bitrate-target codec,simillar to rm9: however,this test show that vp6 looks better at 1,5Mbit/s than rm9...(which isn't a hard thing to acomplish with ANY mpeg4 flavour,even with divx5 beats it..i mean it beats both rm9 AND vp6....),so that means this is not a hi-quality codec,but a lo-bitrate codec,so only place where it can compete is the rm9

if the codec (vp6) looks better than rm9 on 1,5Mbit/s that doesn't mean it'll be better at 300kbit/s....and now we come to the vp6's BIG problem;playback has a serious bug,so when the player (players i tried so far->mpc,wmp6.4,zplayer) window is stretched to anything more than "normal" resolution,the playback starts to drop frames...(if you know a difference between the overlayed and non-overlayed video,then you'll understand it...) (vp6 version i used was 6.0.7.3)
this is,then,unusable video.....untill this bug is fixed,i won't do more vp6 tests.......if and when they fix it,i might do few more tests on ultra-lo bitrates with rm9 and vp6 (i already know rm9 will be much faster on encoding....)..probably 3xx x 2xx @300kbit/s or so....

[update1;or perhaps i'll "fix" it myself?it seems to be working ok (overlay initializes) in tcmp and bsplayer.......addition of rm9 vs vp6 on very lo bitrates seems more likely now]

and here it goes now, rv9 vs. vp6 on 350kbit/s;

the source is here resized to 384x288[bicubicresize(384,288,0,0.75) ] and some sharpening was done with avs' UnsharpMask(strength=10, radius=3, threshold=0)..pretty bad news for vp6 huh....same codec settings for vp6 as on 1,5Mbit/s test,but with bitrate lowered to 350kbit/s.....rm9 encoded the sequence MUCH faster than vp6 too.....

you can see that i wasn't using hypersnap for vp6 screenshots....(remember the overlay/non-overlay issue..seems that when i freeze the vp6 playback also non-overlay mode is entered,so hypersnap can't capture the video screen only,but captures complete desktop....)
but rest assured,i have compared such produced images (hypersnap+vp6+desktop) to vfw(captured from vdub) version and there were no visible diferences,so you're seeing the vdub's <ctrl+1> captured image on above gif image......

to end this lil excursion into video codec field in 2003. , let's review some interesting facts;

bang for your buck (best encoding speed/quality ratio/hi-quality image) category is won by nandub/ffvfw duo....(it seems like nandub should perhaps still be preffered for lower bitrates,while ffvfw clearly won that 1,5Mbit/s test.....anyhow,differences are very,VERY small....both are very nice codecs....although vdub will bitch if you try to load ffvfw with b-frames into it...but that stuff is for watching,and not for editing,right?hihihi)

lo-bitrate star is without a doubt a rv9...this should be the codec of choice for toon encoding folks....now that you can put it into mkv (matroska) files ,then even more so....
(and a way to use ffdshow to post-process it was found too.........
http://forum.inmatrix.com/ikonboard/ikonboard.cgi?s=3f70ad08413effff;act=ST;f=7;t=2387
also,it seems like "tcmp" (the "beefy" one...) will insert ffdshow (if ffdshow is in "raw video" mode...selection is made in "codecs" tab) without any hacks or workarounds...."tcmp" is a nice player......),will do some tests to see if vp6 beats rm9 on very-lo-bitrates.....(done)

mpeg4 codecs will bitch at you,will bite your head of if you wanna spare some space on cartoons.....on cca. 300kbit/s mpeg4 toons will probably block like hell(you would need some mean temporal/spatial smoothing to prevent it from doing so)....rv9 won't........you can even use it as smoothing filter,as rv9 blurs so much(heheh..that's why it's pretty bad for natural content stuff...for lobitrate natural content type of sources, try wmv9)...no big need for elaborate filtering schemes with rv9....

mpeg2 is a codec for interlaced/higher bitrates(above 2,5Mbit/s) stuff.....("above 2,5Mbit/s" is to suggest that svcd's are RULED-OUT for hi-quality encodings...480x576(svcd) usually looks sligthly worse than 576x432(divx)...and if you love interlacing,go that step further....720x576 and the fun starts somewhere above the 2,5Mbit/s....try 3,4 or 5....CCE's VBR mode will do fine...)

i'm done here.........

08.11.2003. by i4004

update1[08.11.2003]-vp6 overlayed playback works with SOME players (tmcp and bsplayer)

update2[08.11.2003]-rv9-vs-vp6 on 350kbit/s added....

update3[09.11.2003]-"IMPORTANT NOTICE" added

update4[10.09.2004]-rv9 is better than wmv9 on lo-bitrates...but only on toon content; on natural content, wmv9 is nicer