PDA

View Full Version : What the best MPEG-2 Software Encoder?


Gabrielgoc
28th August 2002, 02:24
What the best MPEG-2 Software Encoder
a) What is the best quality MPEG-2 software encoder?

1) Ligos
2) CCE
3) Canopus ProCoder
4) TMPGEnc Plus
5) bbMPEG
6) Honestech
7) MainConcept
8) DVMPEG
9) Nanocosmos
10) Other

b) What is the fastest, good quality MPEG-2 software encoder?

(Same list as above)

c) What is the fastest, good quality MPEG-2 Adobe Premiere Plugin?

(Same list if applicable)

THX

vortexer
28th August 2002, 10:17
I think it's always b)CinemaCraft encoder!
you can find some tests on www.chip.de

paulusz
28th August 2002, 10:32
1. CCE SP
2. ProCoder
3. TMPGE
4. Ligos
Forget the rest...

auenf
28th August 2002, 11:29
http://www.tecoltd.com/enctest/enctest.htm

which says tmpg then CCE, with test files available to download too.

Enf...

Arky
30th August 2002, 02:33
YUV (for example miniDV Firewire captures)

RGB (for example MPEG2 DVD VOB rips)


These are two different video standards, and I always find it incredible that people seem to gloss over the fact that it MATTERS that an encoder creates a different output, depending on its INPUT. Any discussion of the *absolute* merits of various encoders, without reference to what source type is being encoded, is surely virtually redundant.

A quote from another forum, recently:

**********************************************************************
"...discussion (of which encoder is 'best') is ABSURD without acknowledgment that different MPEG encoders perform **DIFFERENTLY** depending on whether they are fed with YUV, or with RGB source footage. CCE has an excellent reputation, and indeed it IS excellent, but it does *NOT* encode to as high-a-quality as TMPGEnc Plus where YUV footage is used as the source. The reason everyone in the amateur scene raves som much about CCE is that it is fast, it allows good multipass VBR manipulation, and that the VAST MAJORITY of amateurs use it only to re-encode RIPs of material which they did not generate themselves (MPEG2 RIPs from Hollywood DVDs, which are, by definition, RGB colourspace). For amateurs who actually take the time and trouble to generate their own material on DVcam (which is YUV colourspace), TMPGEnc does a better job than CCE, albeit at an encoding- time-penalty. CCE is not as well-optimised for YUV colourspace source files as TMPGEnc is.

In short, there is no such thing as the "perfect" encoder - it all comes down to educating oneself about which encoder it is best to use DEPENDING on the specific circumstances, and the characteristics of the source footage. In the same way that there is no single type of source footage, so there is no single 'best' encoder which will do the best job on every single type of source footage. "Horses For Courses"

CDrZeus.

**********************************************************************
I think that just about sums up my views on the topic.

I will just say that I consider there to be only three encoders in the running for top honours, each being appropriate in different situations - these are CCE, TMPGEnc, and Procoder.


Arky ;o)

Gabrielgoc
30th August 2002, 16:41
Dear Arky,

I agree with you in the sense that INPUT matters in the quality and the selection of an MPEG-2 encoder.

The info from the "another forum" is incorrect:

Profesional media for SDTV (Interlaced) and their Digital formats are YUV, not RGB. This include DVD MPEG-2 format, and DV format. When an amateur converts (rip) a DVD, hes (shes) making it RGB using conversion programs like DVD2AVI or FlasK. This make easy to encode MPEG-4 (DivX, XviD, wmv, etc) for computer monitor watching. But MPEG-2 is YUV intrinsic.

Im using profesional media, obtained capturing from a TV camera to MJPEG (interlaced PAL SDTV, YUV of course) or miniDV or other DV flavor. Later Im editing the assets to make a TV program (Adobe Premiere). Finaly, my intention is to make a DVD, so, I need convert an AVI (MJPEG or DV) file to a MPEG-2 DVD compliant file.

We are producing material every day for broadcast through a Cable TV System. So we need high quality and very high conversion speed, without the costs of a hardware encoder for little offices (in our main office we have a hardware Vela Research, ARGUS Real time encoder).

In that context what is your answer?

Very truly yours, regards,

Gabriel

unplugged
30th August 2002, 16:45
Some very nice Playstation 2 videos are performed with TMPGEnc :) (like Gran Turismo 3...), they looks excellent.
Just to inform about this...

Bigbucks1959
1st September 2002, 10:43
Uhhhhhh
CCE and tmpg both got 9 score.

But CCE was like done in 1/3 the time...
Sorry I will stick with speed and same quality..

Thx,

Phil K.



Originally posted by auenf
http://www.tecoltd.com/enctest/enctest.htm

which says tmpg then CCE, with test files available to download too.

Enf...

auenf
2nd September 2002, 11:34
Originally posted by Bigbucks1959
Uhhhhhh
CCE and tmpg both got 9 score.

But CCE was like done in 1/3 the time...
Sorry I will stick with speed and same quality..

Thx,

Phil K.





have a look at the test files as well, and keep in mind that it was DV source, some other sources may favour different encoders. (as arky mentioned)

Enf...

DrOutro
3rd September 2002, 20:56
So, this all will end like saying... What encoder I should use?

When I'm encoding some (PAL) DVD's so that they will be put onto DVD-R what encoder is the best for this? (best image quality, I have millions of time so time isn't the essence. Picture-quality is.)

Gabrielgoc
3rd September 2002, 21:10
I have NO "millions of time", so picture-quality is the essence, but as soon as possible. Im using DVD as an asset transport media, and each day we produce a TV program, editing it and finally converting it to a DVD. Then we send the DVD to different location for air playing.

So, Im asking: what is the best MPEG-2 encoder, DVD compliant, that offer very good picture-quality ASAP?

The video input is DV-PAL or MJPEG.

DrOutro
3rd September 2002, 21:28
Ok ok :) You got a good reason to encode it fast. So could we get 2 different (maybe?) answers, one for my usage and one for Mr. Gabrielgoc here...

auenf
4th September 2002, 12:27
if you have millions of time, there is a demo of each of them, try them out with a DVD-RW and see which you like.

Enf...

auenf
4th September 2002, 12:29
Originally posted by Gabrielgoc
I have NO "millions of time", so picture-quality is the essence, but as soon as possible. Im using DVD as an asset transport media, and each day we produce a TV program, editing it and finally converting it to a DVD. Then we send the DVD to different location for air playing.

So, Im asking: what is the best MPEG-2 encoder, DVD compliant, that offer very good picture-quality ASAP?

The video input is DV-PAL or MJPEG.

of the three (tmpeg, CCE SP and ProCoder) CCE SP is by far the fastest, but you may even want to look at a hardware encoder in your situation.

Enf...

DrOutro
4th September 2002, 13:42
Ok ok, I liked CCE SP the most, can it do more than 3 pass and will it make the quality better? I've read somewhere that there could be a 5-pass that you could make or something :P

auenf
5th September 2002, 13:53
CCE SP can do up to 9 passes, but whether over 4 is productive remains to be seen.

Enf...

ErMaC
8th September 2002, 12:13
I don't understand how you can argue about TMPGEnc deling with YUV colorspace better since it converts all input to RGB anyways.

This should have been obvious from the DirectX 8.1 YUV fiasco a while back.

Basically, DirectX 8.1 installed a faulty version of msyuv.dll which converted to 8-bit RGB instead of 24-bit RGB when converting YUV colordata. This meant that opening an AVISynth script that was outputting YUV in TMPGEnc gave you a horrible picture.

TMPGEnc always requests RGB colordata. VFAPI plugins themselves can ONLY output RGB colordata. That's why MPEG2DEC was created - it can give YUV colorspace data to an AVIsynth script which can be fed to another encoder as YUV data.

So to say TMPGEnc handles YUV data better is silly - it converts it to RGB and then handles that. And the silliest thing about TMPGEnc dealing with stuff in RGB colorspace is that MPEG2 is YU12 anyways! MPEG2 stores 4:2:0 colorspace (which is a different 4:2:0 colorspace than what PAL DV uses, just to confuse you!) and as such when it's converted to YUV in AVISynth it goes through a slight upconversion (from YU12 to YUY2) but it's far less damaging then the RGB conversion. So TMPGEnc is taking the source and forcing it thru a YU12->RGB->YU12 colorspace conversion, which is just dumb if you ask me.

CCE can take native YUY2 colordata, although sadly this becomes difficult with v2.62 since it doesn't accept AVS files correctly. Thus feeding CCE 2.5 an AVS script for feeding v2.62 a HuffYUV transcode of your AVS file will put you through fewer color conversions and less data loss overall.

And as should have been obvious from my mention of DVD-MPEG2 above, Hollywood stuff is NOT RGB. ANYTHING on a legal DVD that's been encoded at Main Profile @ Main Level is in 4:2:0 YU12 colorspace (MP@High Level is for HDTV and can do 4:2:2 colorspace i.e. full YUY2).

And for your generating your own content bit with DVCam, you're generating 4:1:1 colorspace anyways (in NTSC, in PAL you're generating co-sited 4:2:0, whereas in the standard US MPEG2 implementation they're interleaved vertically with luma) and therefore you're in YUV land, not RGB land. You would be better off sending your data to CCE in upscaled YUY2 colorspace than RGB, and thus CCE will accept an input closer to your source.
This goes out the window when you start using DVCPro50 or Digital-S, but if you're using that kind of equipment you're not going ot be using CCE or TMPGEnc, that's for sure.

In short - colorspace does not matter to TMPGEnc so arguing which one is better based upon colorspace is silly because CCE always wins since it can accept either. Any quality improvements you see in TMPGEnc have nothing to do with colorspace and instead have to do with the encoder's algorithm itself.

DVDGuy99
8th September 2002, 13:02
Over this weekend I did a comparison of CCE 2.64.01.10, TMPGenc 2.58 and Procoder 1.01.35.0 using an 11 minute mini-DV PAL AVI as the source, with a number of different scenes.

My encoding to DVD-R/-RW is always done with mini-DV as the source. I don't do rips, analog captures, conversions, etc, so my experience is just mini-DV source material, edited in Premiere 6/6.5.

Personally, I find mini-DV source material itself to be somewhat "grainy" or tending to have "mosquito noise" especially in poor lighting conditions. This is all hand held vacation movies, so the camera work is definitely amateur. The time taken to encode is secondary, quality is primary.

I always encode at 8mbs, 10bits DCT, GOP = 12. I use mb1's "Interlaced DV" matrix in CCE and TMPG. Highest quality in both TMPG and Procoder. I had to run "changer.exe" on the CCE output to fix the TFF flag. No other filters where used on any of the encoders.

The result viewed on TV is truly subjective. I sat quite close to the TV and watched each of the three results several times.

I found the quality of Procoder to be slightly better. I call it "smoother" - less grainy, less mosquito noise. I mainly rated the quality on these factors, as I find this to be most visually annoying, especially compared to Hollywood DVD's which are usually silky smooth. But, I did find TMPG to be a bit darker too.

Second in quality ... TMPGenc. Last, but not by much CCE. Sitting back on the couch at normal viewing distance all the outputs were very, very similar.

I didn't accurately note the encode times, but roughly on 1ghz PIII they were 18 min CCE, 1.5 hrs Procoder and 3.25 hrs TMPG.

CCE sure is fast ... but for me, I think I'm going to be using Procoder for my mini-DV to DVD-R encodes.

ErMaC
8th September 2002, 13:06
You might want to consider upping CCE's noise-reduction filters which are located in it's advanced dialog boxes (I forget where). These can help eliminate the kind of noise you describe.

Also important is what DV codec you are using. If it's the standard Microsoft DV codec, I believe it only reports back RGB (which is dumb) and therefore you'd get better quality by obtaining one of the commercial DV codecs which can deal with YUV colordata and then feed that to CCE.

Also note that Premiere works in RGB colorspace, so if you use any transitions or anything that requires processing in Premiere, those scenes get converted to RGB, processed, and then recompressed to DV (which is YUV at co-sited 4:2:0, since you're in PAL land). I believe if you use Premiere 6's DV Playback editing option, it is smart enough to simply splice together the original frames without recompression when you are just doing cutting and splicing, but I can't be sure as I don't use Premiere 6 (I use 5.1 with my DV500).

auenf
9th September 2002, 11:40
intersting comparison Doom9, altho i don't agree that the smoother picture should be the best quality, 'as close to the original as possible' should be the better encoder, noise or not.

altho its handy for procoder to remove a bit of noise on encode, if you use a 'clean' source (ie render), will it make the output a little blurry than it should?

Enf...

DVDGuy99
9th September 2002, 13:02
@auenf Maybe you were commenting on my post? Anyway, you are right of course, not smoother picture but closest to the original is the best criteria. I too am wondering if Procoder might have some secret sauce ( blur ) that makes it appear smoother ... after all, they don't tell you what they put in Coca Cola, it just tastes good. ( please no pepsi - coke debates ! )

Since my only source material is amateur mini-DV, I've personally found the two most troublesome issues in converting to MPEG are low light scenes ( grainy ) and trees ( need a lot of bits to process all those leaves ).

Since I use 8mbs - all the encoders are pretty darn good, and the differences are hard to see and very subjective. I'll run some more tests when I have time ( maybe a couple of weeks, sigh ).

Xesdeeni
10th September 2002, 15:16
I've also use miniDV (amateur) footage (NTSC) to test TMPGEnc vs. CCE. I was fitting 1.5 hours on the DVD, so I encoded at 6.25Mb/s. In TMPGEnc, I used 2-pass (old-type), and in CCE, I used 3-pass. Video was frameserved from Premiere via AVISynth for TMPGEnc, and using the CCE plugin.

When things were still, backgrounds with much detail tended to "pulse" in the TMPGEnc encode. I believe this is the I-picture, since it was about 2 times per second. There were also more artifacts on moving things (with hand-held amateur video, this is almost always), and visible "halos" ("mosquitos").

CCE was much better on all points: no pulsing, negligible artifacts, and very minimal "halos." And it encoded in much less time. The quality of this encode approached many commercial DVDs, IMHO.

Xesdeeni

auenf
11th September 2002, 14:31
Originally posted by DVDGuy99
@auenf Maybe you were commenting on my post? Anyway, you are right of course, not smoother picture but closest to the original is the best criteria. I too am wondering if Procoder might have some secret sauce ( blur ) that makes it appear smoother ... after all, they don't tell you what they put in Coca Cola, it just tastes good. ( please no pepsi - coke debates ! )

heh, didnt read your thread before...

other thing i remembered during the day, 8mbit is all fine and good, what about at lower bitrates (4mbit for interlaced is basically rock bottom) ??

Enf...

ulfschack
16th September 2002, 16:13
Some while back when I didn't have a DVD burner I made some pretty extensive test whether to go with TmpgEnc or CCE. Bitrates were at 2600 kbps, that is to say no vbr allowed. The material was (and still is) interlaced DV.

Now, I don't pretend to know all there is about color space conversions. As I have played around with CODECs, filters, frameservers and what have you, things have never remained the same for longer periods of time. But for comparison of encoders of course I used the same prerequisites. Basicallly there was no way in hell I could tweak CCE to match the quality of TmpgEnc for these kind of avis and for this special kind of end result (SVDC). Not with that kind of shaking and panning!! (my wife's even a worse photographer than I am :) )

Now when putting my movies on DVD maybe I should run another series of tests. But I have a hard time believing that I'd enjoy the same conviction as Xesdeeni has.

I have to agree with ErMac that the improvements are simply in the algorithm itself.

cheers

tyee
27th September 2002, 15:59
Anyone tried the new Mainconcept MPEG encoder Vs. CCE??

tyee

Arky
28th September 2002, 02:22
I *BRIEFLY* tried the MainConcept standalone last night, doing a PAL 720x576 DVD-compliant MPEG2 encode from PAL DV-AVI (interlaced) source footage. I must say I was not particularly impressed, regardless of the supposedly "cheap" pricetag. I liked the variety of encoding parameters which were specifiable, but with the end result, I was not particularly enthusiastic. TMPGEnc is much slower, but much higher quality, IMHO. Of course, CCE-SP is faster, but then there's the cost issue...

Interestingly, I've seen a lot of conflicting opinions on the subject of the encoding quality of the MainConcept standalone/plugin. Personally, I do not see any benefit to being faster than TMPGEnc if the final encode is significantly below standard. I suspect that those who recommend the MainConcept encoder have used unrealistically high bitrates. My encodes tend to range between and average of 3500kbps-to-5000kps. I see no reason to exceed this average bitrate, particularly when two hours or more of footage is going to be squeezed onto a DVD-R. If I have a particularly challenging piece of footage, I up the bitrate, but only for the specific scenes in question. Let me be clear that I would be very happy to be shown a way to get a higher-quality result from MainConcept's encoder - it's just that from what I have seen so far (and, admittedly, after very limited testing) I would not hesitate to recommend TMPGEnc or CCE-SP (for those who can afford it) instead.


Arky ;o)

tyee
28th September 2002, 02:58
I tried it last night too, Arky. I did a segment from Gladiator. I have a 100" dlp projector and when I played it back (3kbps average bitrate), I could see background movement, whatever that's called in the encoding world. I'm going to do the same segment with CCE and then compare. I'll let you know.

tyee

tyee
28th September 2002, 05:43
Well, after comparing CCE 2.5 and MC, both frame served via AVISynth, both averaging about 3.5Mbps, CCE appears to be a little sharper but also has more background motion noise (no noise reduction was used in this test with either encoder). MC has a slightly smoother picture with less background noise. I went through every setting MC has, to be sure everything possible was set to 23.976 and progressive (DVD2AVI forced film source). The defaults that came with MC are not set up for the type of encoding we are doing here. I guess I could live with either but CCE is still faster. On my P3-800MHz, MC took about 2.5 real time, whereas CCE took 1.3 real time. More testing is necessary with a variety of sources but MC looks not too bad, and for the price, well worth it.

tyee

Arky
1st October 2002, 01:36
Found this interesting thread on VCDhelp.com. It's so long that I haven't had a chance to read through it all yet, but that doesn't mean you can't!

h**p://www.vcdhelp.com/forum/viewtopic.php?p=331760#331760


Arky ;o)