Comparison of TMPGEnc 2.56 vs. CCE 2.64 vs. Main Concept

[BJ_M]

Source: Digital Beta tape captured w/ Pinnacle ReelTime Nitro capture board w/ component output, MJPEG format 6meg/s 720 x 486 (comparable to DV captured via firewire, i kept bit rate low on capture). combination of animation and live shot. fast motion with high contrast and also white txt on black background (its a good little test sample i made from a project i was working on).

AVI trim'd in V-dub for lenth , no filters. this is a interlaced clip - no pulldown applied.

Encoders:
1. CCE 2.64
2. TMPGEnc 2.56
3. Main Concept standalone 1.01
4. Main Concept Vegas Video 3(and sonic foundry) plug in
5. Ligos 2.0 (not posted results)
6. Optibase Fusion mpeg2 high end pro hardware encoder (not posted results - but it does up to 15meg/s @ real time in VBR mode and has very very good quality - but software encoders can beat it quality wise)
7. Panasonic mpeg1 & 2 "new" encoder (demo)

Quality was checked with a Snell and Wilcox analyzer and compliant streams were checked with M-Probe

Clip Settings:

1. CCE DVD mpeg2 - DVD template, standard matrix, 3 pass VBR, 7000 max , 5000 avg, 2000 min (no pad), 9 on quality setting. no other settings applied.

2. CCE SVCD template mpeg2 - 3 pass VBR , 9 on quality , all other settings default svcd. (sorry didnt resize to 480 x 480)

3. CCE VCD - sucked so bad i didnt post it -- default mpeg1 cbr and vbr both bad quality IMO ..

4. TMPGEnc DVD mpeg2 - default 2 pass DVD setting, 7000 max , 5000 avg, 300 min no pad, default matrix, slowest motion search, do not search by half pel. slowest encoding - rated 2 points behind cce in quality out of 50 max.

5. TMPGEnc SVCD mpeg2 - default SVCD setting , slowest search , 2 pass vbr, "andre's" matrix applied (from scvd2dvd), do not search by half pel. highest rating in quality, slowest in speed.

6. TMPGEnc VCD mpeg2 - default VCD setting completly. ok speed - still slow though, high quality , got same rating as MC but i think color looks better.

7. Main Concept DVD mpeg2 - default dvd setting , VBR mode 7000 max , 5000 avg , no changes to any settings. stand alone encoder - vegas video plug in at best setting was also used and results were very close the same except standalone about 30% faster. Standalone was fastest encoder - beating CCE by almost twice the speed but only doing a 2 pass vbr. created much smaller files size than CCE . 3rd in quality behind tmpgenc and cce but only by 3 points out of 50

8. Main Concept SDVD mpeg2 - default sdvd setting , VBR , no changes to any settings. stand alone encoder , on its own MC changes its matrix for svcd. Fastest encoder - beating CCE by almost twice the speed but only doing a 2 pass vbr. smaller file size. same quality as tmpgenc - slightly better rating for quality than cce.


9. Main Concept VCD(1) - default vcd setting , encodes at CBR in default settings. 3 x real time speed . small file size and good quality - tmpgenc a little better quality wise -- but marginal and with tweaking they would be about the same.

10. Main Concept VCD(VBR) - vcd setting , default settings but i changed it to VBR instead of default CBR for vcd

11. Panasonic VCD (CBR) - default settings

12. Panasonic SVCD (CBR) default settings -- i cant get it to do VBR , may be because in demo mode ?

13. Panasonic DVD (CBR) 7000 - default settings.


Notes on the panasonic encoder - it doesnt do to well on speed or bit rate or quantisation level - see pics below - it also created the largest file sizes (though as i mentioned couldnt get VBR to work). It CAN do very high bit rates - even High def 50meg/s oddly enough and at real high bit rates it was not bad (though slow):

Panasonic DVD bitrate vs. quantisation level (it should not be crossed normally)

Panasonic SVCD bitrate vs. quantisation level

compare this to the CCE DVD levels (MC and TMPGEnc are closer to CCE)


All in all i am very impressed with Main Concept for two reasons - one its is the fastest encoder of this bunch and second it can produce very small file sizes (size on dvd sample seems ultra low and im not sure why - havnt looked into it yet) while keeping quality (extremly good at VCD and low bit rate) .. I used as much as possable default settings instead of a lot of tweaking to more nearly level the playing field - of course one can adjust thier own settings - but also I am trying to show, if you look at the quality of the samples, that often a lot of tweaking can cause more problems than help unless you are really sure of what you are doing.
Downsides to Main Concept is that it would sometimes not encode an entire longer avi segment .. don't know why. also i had a lot of problems with it in settings up custom settings and saving them as a template, both as making up my own templates and also using its save settings feature ..
I hope Main Concept (reads this) fixs these issues asap. They should also have the feature of CCE where i can change the offset of the time code of an encoded file (used by some authoring purposes - tmpgenc doesnt have this at all). I have not explored the ability of Main Concept to use numbered file seq's as TMPGenc can and CCE can not - which is one of my bigest complaints with CCE ' s use on a professional level. .

im not going to get into which is best - people can judge for themselves and also I am well aware of the fact that no mater what - there will be preconceptions.

If anyone wants more info or a comparison with other encoders or a different type of source or capture (i have others also) - let me know.




THE SAMPLES
(OFF LINE Right Now - new samples - soon

1. CCE DVD Sample 10.1 meg

2. CCE SDVD Sample 3.7 meg

3. CCE VCD Sample NOT POSTED 'cus it sux

4. TMPGEnc DVD Sample 9.7meg

5. TMPGEnc SVCD Sample 3.8meg

6. TMPGEnc VCD Sample 2.6meg

7. Main Concept DVD Sample 6.4meg

8. Main Concept SVCD Sample 3.8meg

9. Main Concept VCD default settings Sample 2.6 meg

10. Main Concept VCD set to VBR 1.9meg

11. Panasonic VCD

12. Panasonic SVCD

13. Panasonic DVD - not posted but see above..

[SLK001]

BJ_M,
Your pointers to your Main Concept DVD clip points to a TMPG clip.

[BJ_M]

thanks - fixed

[TheSeven7hSon]

Sample #8 points to TMPG.

Wow, what a difference between CCE and TMPG on the SVCD shapness of the lettering.

Seven

[Lamont Cranston]

Downsides to Main Concept is that it would sometimes not encode an entire longer avi segment .. don't know why

MainConcept V1.1 only supports openDML for dv type 1 avi files which could be the reason you could not encode the full length (<4G). I am not sure if this was the cause in your case as you did not mention your file type. THe MainConcept encoder will crash if you try to encode video only. If you choose mpeg 1 layer 2 audio you have to go to the advanced setting and reselect otherwise you will get a long list of errors. I am assured these will all be corrected in a later version.

I looked at the MainConcept's encoder strictly for DVDs. The test clip I used for evaluting mpeg encoder showed a lot more artifacts with the MainConcept Encoder that with TMPGEnc. In one test I ran TEMPGEnc at a video data rate where I just started to see blocking in some areas. With the MainConcept encoder running under similiar conditions I did not see blocking but a latent image only visible when the scene moved. As the image panned over a unifomly illuminated and colored surface you could see the latent image or artifacts moving in the background but no blocking. I don'y know what they are doing. My tests are very subjective and not quantative.

One of the features I like is being able to encode multiple avi files into one mpeg file. This gives me the option to have different video data rates for each segement. I have some segments that reguire a 8K video data rate becuase of the action and others where an avg of 4K would be fine. Helps keep the file size down without scraficing quality.

[BJ_M]

Sample #8 points to TMPG.

Wow, what a difference between CCE and TMPG on the SVCD shapness of the lettering.

Seven

link fixed - sorry

[BJ_M]

Downsides to Main Concept is that it would sometimes not encode an entire longer avi segment .. don't know why

MainConcept V1.1 only supports openDML for dv type 1 avi files which could be the reason you could not encode the full length (<4G). I am not sure if this was the cause in your case as you did not mention your file type. THe MainConcept encoder will crash if you try to encode video only. If you choose mpeg 1 layer 2 audio you have to go to the advanced setting and reselect otherwise you will get a long list of errors. I am assured these will all be corrected in a later version.

................

i was using a huffyuv compressed 3.5gig file in that case -- tried again with same file but using mjpeg compression and same problem - same lenth , may be a frame it didnt like but cce and tmpgenc didnt have a problem ...

had no problems encoding video only (i always encode video only) - though i will keep the tip in mind if i have a problem .. perhaps that was same reason why had problem above as to lenth -- no problems on long files on plug in in Vegas Video (had encoded several 1 hour segmesnt in vegas video all with out audio (audio saved out as pcm to be encoded to ac3)).

In-Sync Blade uses the TMPGEnc mpeg encoder btw if anyone is interisted - i am evaluating that software over the next week or so .. it is a lot like Vegas Video .. Blade is the first program i know of that uses TMPGEnc as a encoder ..

[BJ_M]

i have found that Main Concept will always encode the whole file proper if i dont input 30fps captures and encode them at 29.97 .. just feeding the capture through v-dub set to convert frame rate will fix the problem ..

also make sure (as you noticed) that you set video drop down box to video only , not program if you are not using audio stream , then i get the problems you mentioned before ....

[gf]

Greetings BJ_M
I was looking at your video's and they were GREAT. Could you please tell me a little more about your system ex. capture card,motherboard,op sys...

PLEASE

Thanks
Steve

[BJ_M]

Greetings BJ_M
I was looking at your video's and they were GREAT. Could you please tell me a little more about your system ex. capture card,motherboard,op sys...

PLEASE

Thanks
Steve

look at my system (click icon named "computer details") but for those samples I used a Digital Betacam 510 - captured with a PINNACLE ReelTime Nitro card (this a OLD card from 1999) with the component inputs (Not SDI for this test). capture details in first post above.

i captured to a WD 120gig drive (there are 6 - 12 x 60 -120 gig drives in this system at any one time) , encoding in this case was done on a p4 1.8 and an epox MB with 700 something memory - only 133 sdram memory. video card is a Xabre400Q 128meg wth dual monitors (2 x sony G500) --- this is my capture only machine for the most part as well as part of a render node. Nothing special really other than it also is attached to the DLT tape system. shares the monitors with other systems .. I use win2k sp3 on that machine. it is a real ugly thing as im always changing drives.

[Thorn]

Thanks for the comparison, BJ_M. The clips and analysis are quite helpful... my eyes go blurry after watching clips too long so I compared the text and then relied on your comments . It seems that CCE handled the text the best by far, but the video looked good on all. Anyway, this kind of info should be in the guides (I don't think they've updated the comparison here for a while).

Also - nice workstation. Slightly upgraded from the standard consumer's video editing rig...

[vhelp]

my pc looked like your first pic, but I've since, narrowed down the bulk of junk
hanging out of the opened case as in yours.

With all that junk, I'm surprised you don't have any Line Noise in your
samples, as I had in some of mine.

Ok, NOW do a VHS sample, he, he...

-vhelp

[Lamont Cranston]

i have found that Main Concept will always encode the whole file proper if i dont input 30fps captures and encode them at 29.97 .. just feeding the capture through v-dub set to convert frame rate will fix the problem ..

Were you encoded your sample did you use the stand-alone or the plug-in MainConcept encoder? I used the stand-alone. Here is MainConcept's tech support email response:

you are right. Special applications are now able to capture DV-type 2 files with openDML. We are working on this feature for the encoder. Large DirectShow files should be possible to be imported. Is it possible to capture in type-1? I do not know if the VFW OpenDML support will be available in the next released version.

All the problems I have encoutered are 'front end' problems and have nothing to do with the encoder. IMHO The encoder does a good job.

BJ_M Your samples look very good! When you ran TMPGEnc DVD test did you look at pass 1 of the encode log? If yes, what was the range of the bit rates?

[fatdom]

Not wanting to appear stupid but how do you change from 1xVBR to 2XVBR in the mainconcept standalone (1.1). Cheers, Dom.

[Lamont Cranston]

Not wanting to appear stupid but how do you change from 1xVBR to 2XVBR in the mainconcept standalone (1.1). Cheers, Dom.

First, there are no stupid questions here. On the first screen open your file then click on the Details button, then the Advanced button. On the lower right you will see the maxiimum, average and minimum bit rate. VBR can only be used with SVCDs and DVDs.

[BJ_M]

fatdom- to get the VBR sample i did above for vcd i had to select Mpeg1 instead of VCD andf then adjusted settings to match vcd - which pretty well i recall were already there as default.


Lamont Cranston - samples above are from stand alone Main Concept and are almost the same results using plug in. Yes i agree - the encoder does a VERY good job and can produce smaller file sizes than other products. And yes - it was a front end problem in that case - though its not holding its settings and not also applying some changed settings correctly are a problem still.


the custom svcd matrix table i used for TMPGEnc is below -- you can just add this to your TMPGEnc ini file ...

Name3="custom_svcd"
ReadOnly3=0
Intra3_0=8 16 19 22 26 27 29 34
Intra3_1=16 16 22 24 27 29 34 37
Intra3_2=19 22 26 27 29 34 34 38
Intra3_3=22 22 26 27 29 34 37 40
Intra3_4=22 26 27 29 32 35 40 48
Intra3_5=26 27 29 32 35 40 48 58
Intra3_6=26 27 29 34 38 46 56 69
Intra3_7=27 29 35 38 46 56 69 83
NonIntra3_0=16 17 18 19 20 21 22 23
NonIntra3_1=17 18 19 20 21 22 23 24
NonIntra3_2=18 19 20 21 22 23 24 99
NonIntra3_3=19 20 21 22 23 24 99 99
NonIntra3_4=20 21 22 23 25 99 99 99
NonIntra3_5=21 22 23 24 99 99 99 30
NonIntra3_6=22 23 24 99 99 99 30 99
NonIntra3_7=23 24 99 99 99 99 99 99

[Lamont Cranston]

BJ_M The samples looked so good I download the DVD test samples for each encoder and burned a DVD so I could look at them on the big screen.

For some reason Ulead's DVDWS reported the maximum video data rate different for Mainconcept and CCE from those you posted.

MainConcept DVD Max Video data rate 6000kbps vs 7000kps
CCE DVD Max Video Data rate 8500kbps vs 7000kbps

They all looked good and essentially the same as your data suggests. I clicked on the Sony Video Data Rate monitor and what was interesting is the video data rate for both the CCE and TMPGEnc ranged around 6000 -7000 kbps and 4000kbps - 5000kbps for MainConcept during the maximum action. The minimum data rate for the MainConcept was also about 1000kbps - 2000kbps lower that tmpgenc and cce. Yet there was no apparent quality difference in the picture. From tests that I have run where I lower the video data rate for both TMPGEnc and MainConcept until the video starts to degrade TMPGEnc starts to degrade at a much higher video data rate than MainConcept. What is MainConcept doing?

[BJ_M]

ODD -- i did set all to 7000 max .. but i certainly believe it to be the case..

main concept has additonal templates on the adobe 6.5 cd also ...

and now Main Concept has lowered their price to $49 from 149 (if you have adobe 6.5) which includes the main concept dv codec (no i dont work for them!) which makes it a pretty cheap solution i wouldnt hesitate to suggest over a $2000 solution (CCE) -- and yes it work with frame serving and all types of input files (seq tga files need to be encoded using Main concept as a plug in).

[Lamont Cranston]

BJ_MWhen you ran TMPGEnc DVD test did you look at pass 1 of the encode log? If yes, what was the range of the bit rates on your test clip?

[fatdom]

Yeah, I saw the checkbox for VBR and the min, max and average but isn't that just one pass VBR. I couldn't find anyway to do a two pass VBR. Although to be honest the one pass encodes are top notch anyway, cheers Dom (still fat)

[snowmoon]

BJ_M:

Would those same matrixes work under CCE? I understand most of the numbers ( in a general sort of way ). Why those two lost 30's in the non-intra matrix?

[BJ_M]

BJ_M:

Would those same matrixes work under CCE? I understand most of the numbers ( in a general sort of way ). Why those two lost 30's in the non-intra matrix?

yes - it works with cce also -- in fact i use cce matrix in tmpgenc often ..

the two lost 30's -- prob my error (hehehe) well it works - should be 99's, i missed that ..


Lamont Cranston - i didnt look --- i will redo that one and check ..

[BJ_M]

Lamont Cranston - this is first pass log

---- 1 pass (number of pass attempted)
frame=       0 ( 16.0)  Bitrate: 1418 kbps  Motion:  0.36  MotionScale:  0.00  CQ: 58.89
frame=      16 ( 16.0)  Bitrate: 1236 kbps  Motion:  0.34  MotionScale:  0.00  CQ: 58.89
frame=      32 ( 16.0)  Bitrate: 1347 kbps  Motion:  0.43  MotionScale:  0.00  CQ: 58.89
frame=      48 ( 16.0)  Bitrate: 1220 kbps  Motion:  0.34  MotionScale:  0.00  CQ: 58.89
frame=      64 ( 16.0)  Bitrate: 1215 kbps  Motion:  0.34  MotionScale:  0.00  CQ: 58.89
frame=      80 ( 16.0)  Bitrate: 1250 kbps  Motion:  0.34  MotionScale:  0.00  CQ: 58.89
frame=      96 ( 16.0)  Bitrate: 1352 kbps  Motion:  0.38  MotionScale:  0.00  CQ: 58.89
frame=     112 ( 16.0)  Bitrate: 2876 kbps  Motion:  1.95  MotionScale:  0.00  CQ: 58.89
frame=     128 ( 16.0)  Bitrate: 2697 kbps  Motion:  1.45  MotionScale:  0.00  CQ: 58.89
frame=     144 ( 16.0)  Bitrate: 1753 kbps  Motion:  0.53  MotionScale:  0.00  CQ: 58.89
frame=     160 ( 16.0)  Bitrate: 2297 kbps  Motion:  0.88  MotionScale:  0.00  CQ: 58.89
frame=     176 ( 16.0)  Bitrate: 2265 kbps  Motion:  1.12  MotionScale:  0.00  CQ: 58.89
frame=     192 ( 16.0)  Bitrate: 2428 kbps  Motion:  1.00  MotionScale:  0.00  CQ: 58.89
frame=     208 ( 16.0)  Bitrate: 2896 kbps  Motion:  1.95  MotionScale:  0.00  CQ: 58.89
frame=     224 ( 16.0)  Bitrate: 2660 kbps  Motion:  1.63  MotionScale:  0.00  CQ: 58.89
frame=     240 ( 16.0)  Bitrate: 3674 kbps  Motion:  2.07  MotionScale:  0.00  CQ: 58.89
frame=     256 ( 16.0)  Bitrate: 4135 kbps  Motion:  4.43  MotionScale:  0.00  CQ: 58.89
frame=     272 ( 16.0)  Bitrate: 3513 kbps  Motion:  2.79  MotionScale:  0.00  CQ: 58.89
frame=     288 ( 16.0)  Bitrate: 4192 kbps  Motion:  4.17  MotionScale:  0.00  CQ: 58.89
frame=     304 ( 16.0)  Bitrate: 3228 kbps  Motion:  2.38  MotionScale:  0.00  CQ: 58.89
frame=     320 ( 16.0)  Bitrate: 2575 kbps  Motion:  1.43  MotionScale:  0.00  CQ: 58.89
frame=     336 ( 16.0)  Bitrate: 2310 kbps  Motion:  0.96  MotionScale:  0.00  CQ: 58.89
frame=     352 ( 16.0)  Bitrate: 1686 kbps  Motion:  0.66  MotionScale:  0.00  CQ: 58.89
frame=     368 ( 16.0)  Bitrate: 1871 kbps  Motion:  0.76  MotionScale:  0.00  CQ: 58.89
frame=     384 ( 16.0)  Bitrate: 1625 kbps  Motion:  0.55  MotionScale:  0.00  CQ: 58.89
frame=     400 ( 16.0)  Bitrate: 1791 kbps  Motion:  0.70  MotionScale:  0.00  CQ: 58.89
frame=     416 ( 16.0)  Bitrate: 1877 kbps  Motion:  0.81  MotionScale:  0.00  CQ: 58.89
frame=     432 ( 16.0)  Bitrate: 1464 kbps  Motion:  0.41  MotionScale:  0.00  CQ: 58.89
frame=     448 ( 15.0)  Bitrate: 1197 kbps  Motion:  0.32  MotionScale:  0.00  CQ: 58.89
Surveyed bitrate : 2211 kbps    average CQ: 58.89
---- 2 pass (number of pass attempted)

[Lamont Cranston]

Thanks BJ_M.

[ABS]

I have just noticed something...
The video you encoded isn't interlaced, but you encoded it as one... True that it has 480 lines, but I don't see any interlace... (I look in a program that doesn't do de-interlace).

Also, the text at the begining on TMPEGenc seems to have artifact caused by that fact it was encoded as interlace (maybe more sensitive then other programs?)

I think you should try to encode as progressive and see if there is any difference...

[BJ_M]

it is a interlaced source and capture for sure .. no question about it .. since i made it ..

[ABS]

Do you have the source file or is it too big to download?

[BJ_M]

sure i have the source files - as well as the tapes and even the original animation frames (i didnt do the animation, its from a movie though ) - but the source for this sample was digital beta cap'd through a card i have been using for several years --i certainly know interlaced vs. non interlaced .. that sample i seem to recall was about 400meg , since its copyrighted materal im not going to send out the source files even though i watermarked them. i can get some interlaced lines like anyone (but its its being played back on a tv its not the end of the world) - but really i am trying to point out not only the qualities of various enocders - but the fact is even at basic standard template settings you can get very good mpeg's ....

the biggest factor in quality of your compression is the source material... and the fact of keeping highest color settings throughout all steps and avoiding things like making a 720 x 480 mpeg2 from 720 x 486 cap (this is a bad thing if you do a simple resize on this combination) and other factors of good capturing practices ..
I cant say enough to use good cables for both he audio and video - have all video equipment on same power connections, use good connectors with proper impeadances that are tight and clean, if you dont have really good cables - at least keep them short. i use long cables - which is not a factor if you have quality cabling.
make sure heads are clean on your vcr -
it is better to make a 1/2 D1 cap at a higher data rate at max color than to make a full D1 cap at a lower rate.
avoid re-compressing your material - i see people capturing with mpeg then converting to divx ..
if you have noisy sources - sometimes capturing in a lossy format like some mjpegs can accualy improve the quality.
aviod procccing your video - yes keep any proccessing to the min as almost any type of filters or size changes or whatever will often degrade the quality more than help it .. i see people (a few) using big numbers on this forum for noise reduction and im not sure why as that kills the picture-

[MNale]

Please explain...

"avoiding things like making a 720 x 480 mpeg2 from 720 x 486 cap"

I capture from DigiBeta to uncompressed 4:2:2. My SDI capture card, like yours, defaults to 720x486. How else can one make a DVD compliant MPEG2 unless it is resized to 720x480 while encoding?

[snowmoon]

BJ_M:

Would you qualify this as too much NR. I have a noisy calbe feed ( probably average for most people ). I belive my NR is light to moderate as I find the smearing and ghosting from over NR to be more objectionable than the noise in the first place.

I cap blacks to 16 and lights to 235 no other leveling is necessary because I have calibrated my brightness and contrast to begin with. I do a 2x2 smoother to restore continuious tone areas ( it does not smooth edges ). Then I do a temporal filter that has a ramp based on luma channel ( ie the brighter it is the more leway it has to filter ). The temporal is a non-locking, if two pixels are close enough to be filtered it blends them together. At this point I do a bicibic resize from 720x480->480x480 and in some cases I do a very light sharpen to bring out the pop again. I only do sharpening on progressive materal since I don't have a good interlaced shapeneing filter.

If you want to see some before and after it really does make a huge diffrence in quality of output materal.