Searching for the best low-bitrate mpeg2-softwareencoder

[peteag]

I want to author a full-featured dvd with everything you could imagine but my problem is: what mpeg2-encoder is the best at the lowest possible bitrate!!! But everywhere I heard this and this. This little poll may help me and you by just choosing your favorite encoder.

Thank you.

[Fishman0919]

Forum Rule 12

12) How NOT to post on this forum:

5 minutes after your initial post you reply to your own post writing something like "Why is nobody helping me? I'm sure you know the answer". Once somebody replies to your thread but doesn't give you exactly the answer you've been expecting you insult them.

Do not ask "what's best" because this question cannot be answered objectively. Each and everyone has their own view about what's best in a certain area. The best is what works best for you!




Maybe you could ask "What is everyones favorite encoder at low bitrates"

[Encoder Master]

The best for DVDs is at the moment HC Encoder. I hope hank can also in the future improve more quality and speed with HT encoding.

CCE und mainconcept aren't very good a low bitrates. FreeEnc and QuEnc give good quality, too.

TMPGEnc is the best at low bitrates for one-cd in MPEG-1.

[peteag]

i've done some encoding sessions last night and if i could i would uncheck the procoder instantly. my source is a very dirty (simulated dust, noise, scratches etc.) b/w progressive sequence of 3000 frames, 25fps (PAL 720x576), aspect 16:9. i've tested cce 2.70.02.00, procoder 2 and hc 0.14 at 0min/3200abr/4800max. cce did the best job without destroying the sharpness of the noise and not that much blocks. procoder did almost the same but the image was something smoothed but also no blocks at all. hc did a good job too but there were more blocks visble.

sidenode. procoder and hc put the picture 1 pixel down. why this?

i need to encode anything at the lowest possible bitrate because it's a lot of material to be burned on the dvd ;-)

[Archimedes]

Beside the subjective tests you can also make some objective tests with
http://forum.doom9.org/showthread.php?s=&threadid=93403

Hope it helps.

I assume that your videos was encoded with top field first. Take a look at your CCE settings. There is an option called offset line, which is default set to 1. This is the reason why your encodet CCE video is shifted by 1 line (CCE's solution for top field first). Procoder does it in the correct way.

[kolak]

"Best encoder" in low bitrates is Sony Vizaro hardware encoder:)

[Koepi]

I prefer QuEnc. It simply rocks at low bitrates. It's even better now with XviD's ratecontrol. No more buffer over/underflows and stable quality.

But as usual, this one seems to be missing.

[SeeMoreDigital]

Hi peteag,

What sort of "lowest possible" bit-rate are you considering?


Cheers

[peteag]

lowest possible: 3-4mbps !!!

[Encoder Master]

I prefer QuEnc. It simply rocks at low bitrates. It's even better now with XviD's ratecontrol. No more buffer over/underflows and stable quality.

But also with XviD Rate Control I have some too high peaks.:mad:

lowest possible: 3-4mbps !!!

This is a realy high bitrate would I say. Then you can use a lot encoders which gives you very good quality. HC Encoder and CCE seems to be your best choise because they are very stable and gives you very good results. I though you mean an average Bitrate between 1500 and 2000Kbps with a res of 720- or 704x576 then I use at the moment HCE.

[peteag]

Yes, this is possible? 720x576 PAL 25fps at 1500-2000mbps - very good quality? this is possible?????

but why hc? what makes him so special ?

i'll do some testings right now ;-)))))

hints to my material:

- black/white or color
- simulated (realy good) film damage with flickering effects but also normal footage
- 2.40:1 (16:9)
- high motion

[Encoder Master]

If it's a 16:9 Letterbox movie with a good compression matrix and it depend on the movie then it is possible. HC Encoder has a very good motion search.

[peteag]

no, its very important for me, that the movie is in 16:9 (2.40:1). what means that the matrix depends on the movie. how to create/get such a matrix? what is the best matrix for very very low bitrate on hc?

[SeeMoreDigital]

720x576 Mpeg1 @ 2000Kbps can look very good indeed... But I don't know whether the DVD specs allow this resolution!


Cheers

[peteag]

what about dvd-specs. when i encode a black screen, with just a simple title in the center, the bitrate don't expires 1mbps - 1,5mps. and this can be decoded by a dvd-player very well. look at some dvd-titles and their opening-titles. the bitrate is as low as 1-1.5mbps. what's the problem?

[Encoder Master]

no, its very important for me, that the movie is in 16:9 (2.40:1). what means that the matrix depends on the movie. how to create/get such a matrix? what is the best matrix for very very low bitrate on hc?

I can only say what I use if the bitrate on my DVDs is too low. Then I use the MDVDex matrix which is very good for DVDs and the quality is great. And also with 16:9 (But in 2.40:1 like your source).

what means that the matrix depends on the movie

This means: Is it high motion or not or dark or something like this. An example. I put the whole Matrix - The first movie on one CD and the quality looks great because there are so many dark scenes which don't use a lot of bits. But other movies are harder to compress.

720x576 Mpeg1 @ 2000Kbps can look very good indeed... But I don't know whether the DVD specs allow this resolution!

No, only 352x288 is DVD conform and a max bitrate of 1800Kbps. In MPEG-2 and a good encoder it looks good.

[Mug Funky]

procoder and hc put the picture 1 pixel down. why this?

actually, it's CCE that's moving the picture 1 pixel up (bear this in mind when doing PSNR style comparisons as this will mess things up unless you move the image down by a pixel).

CCE will always output top-field first. if it's fed with bottom-field first, it'll simply move it up 1 pixel to reverse the field dominance and encode as TFF. to stop this, just tell CCE your source is top-first to start with (it's fine to do this on progressive sources, but telecine/interlaced sources will judder).

as for bitrate, nothing will look good at 2000 except a talking head on a black background. as far as minimum bitrates that a DVD player can handle, it depends on the make/model. a rule of thumb is "if it plays on a samsung, it plays on everything" - this is not because of strict DVD compliance, but simply because samsungs tend to have more difficulty than other brands for some reason.

@ kolak:

would an MPX-3000 do? they kinda suck below 4mbps though.

[Revgen]

I prefer QuEnc. It simply rocks at low bitrates. It's even better now with XviD's ratecontrol. No more buffer over/underflows and stable quality.

But as usual, this one seems to be missing.

I like to use Quenc for low-bitrate cartoon and anime encodes. Especially with the QLB matrix.

But for live-action encodes at low-bitrates I prefer HC with the AutoQ2 matrix.

[Kika]

The best Encoder for low bitrates in MPEG2? TMPGEnc! And don't even think about this MPEG1-Rumour. TMPGEnc is better in MPEG2 than in MPEG1-Encoding - if you know, how it works. ;)
HC isnt' good at low bitrates, because it does not vary the quantisation enough.
QEnc also isn't good, becaus of it's bad calculation for Bitrates.

With TMPGEnc, you are able to set EVERY Parameter, but no one is doing this, it's a pity.

I think, HC is a good candidate for becoming the winner in this type of "contest", but hank has to implement some special settings like vary quantisation if the bitrate is too low and a setting for the Min.-Bitrate in Constant Quantisation Mode.

[SAPSTAR]

My favorite encoders are :
CCE SP 2.70, HC, AQE, TMPGEnc depending on what I'm encoding ....there is no BEST encoder...they all have their strength in particular areas.....

Personnaly, for ultra low bitrates I think AQE is better than others...but I may be biased! ;)

[Kika]

*g*
But to be serious: The key to a good result ist to understand, how an encoder works - to know it's advantages and it's weakness.
A dynamic change of the Matrix is a interesting Idea, but Matrix changes do need more bitrate (for the Matrix). And i prefer do use bitrate for the picture, not for anything other.

[Fishman0919]

*g*
But to be serious: The key to a good result ist to understand, how an encoder works - to know it's advantages and it's weakness.
A dynamic change of the Matrix is a interesting Idea, but Matrix changes do need more bitrate (for the Matrix). And i prefer do use bitrate for the picture, not for anything other.

ah.. But isn't understand how something works the key to knowing it's advantages and it's weakness and using them to your advantage... without truly knowing how something works, you will never really understand it.

but Matrix changes do need more bitrate (for the Matrix).

:confused: I think you might not understand how a matrix is used in an encoder. Depending on the matirx, it's possible to uses a lower bitrate and get the same quality because of better compression from an optimized matrix.

[SAPSTAR]

*g*
But to be serious: The key to a good result ist to understand, how an encoder works - to know it's advantages and it's weakness.
A dynamic change of the Matrix is a interesting Idea, but Matrix changes do need more bitrate (for the Matrix). And i prefer do use bitrate for the picture, not for anything other.
I think you didn't understand the way AQE works...it calculates the best matrix for the vid it's encoding...but it does NOT chnage it "dynamically" inside the vid itself, it could cause some issues on some players in addition to some other pbs....
So by having the optimum matrix for the movie, you can achieve a better usage of the bitrate for a better visual quality !
With DVDRB, it's rather interesting because the movie is splitted in several segments so each segment has its own matrix......

[Kika]

Oh, ok, my mistake.
But what are the rules for calculating the "best matrix"? Best Picture quality? Best compression?

[gigah72]

some links for reading about dctune
http://forum.doom9.org/showthread.php?t=34034&highlight=dctune

[SAPSTAR]

Oh, ok, my mistake.
But what are the rules for calculating the "best matrix"? Best Picture quality? Best compression?
Well it's a bit complicated, simply, I developped two formulas based on the bitrate and darkness of the current scene to pass arguments to dctune2.0, I apply this treatment to an I-pic and a delta pic every two GOPS, then the resulting matrix is averaged with the partial results coming the previous GOPs.....the final result is then MPEG2 normalized by moves. Then this matrix is used in the encoder engine....
So the matrix is supposed to give the best visual results for the given bitrate.

[Delphin]

I just ran some test encodes of old television shows (shot on film 23.97 fps) and have to say I am very impressed with the free HC MPEG2 encoder (now using version 15).

If you need REALLY LOW bit rates try using AVISYNTH to downsize the video to NTSC CIF resolution (352x240) and then encode at a bit rate of only 750kb Average / 2000 peak using HC with 8 bit DCT and the JAWOR1CD matrix.

I did this, then authored the resulting MPEG2 video with DVDAUTHOR and the resulting video, when played back on my standalone Philips DVD player, could easily be mistaken for a standard DVD D1 encode when viewed from across the room on a standard 30 inch screen.

There is no visible blocking, mosquito noise or any other digital video artifacts (it looks very close to a good quality analog VHS tape).

The same settings could be extrapolated to a 720x480 NTSC D1 encode by simply raising the data rates by a factor of 4 to an average of 3000 with a peak of 8000 (and still using 8 bit DCT accuracy setting and the JAWOR1CD matrix).

I understand that these aren't the absolute lowest possible bit rates, but they were amazingly low rates for such a high quality encode with NO ARTIFACTS.

This points out one of the reasons that there is no 'BEST' answer to your question, because each encoder works a little differently and, more importantly, people have dramaticaly different tollerences to the 'digital artifacts' which happen when you push things too far.

So which encoder 'wins' depends on what criteria you use (speed/quality/etc) and even if we try to nerrow things down to just 'quality' it's not that simple because it depends on where you draw the line.

For example HC is very very good at creating seamless 'comercial quality' encodes at moderate bit rates like I mentioned above, but can get a little nasty looking at extreamly low 'webcam' like bit rates. On the other hand QuEnc gets pretty blocky at very low bit rates but may (to some eyes) remain more 'watchable'.

Here are some general observations regardless of the encoder you pick . . .

8 bit DCT accuracy is usually better for low bit rates (vs. 9 or 10 bits which are better for high bit rate encodes)

Matrices which have a greater range from upper left to lower right are tailored for higher compression and lower resulting bit rates (try JAWOR1CD, BACH1, QLB etc.)

24fps (23.97) progressive frame material can be encoded to lower bit rates than 29.97 interlace (both the higher frame rate and interlace each cost bits), so for NTSC encoded films, always try to remove pulldown frames and de-interlace (if required) BEFORE encoding to restore the original 23.97 progressive characteristics of the film.

As a last ditch, putting a 16:9 widescreen film in a 3:4 frame with black ‘letterbox’ boarders BEFORE encoding can dramatically lower the required bit rate vs. a true ‘anamorphic’ 16:9 encode. This happens because the black boarders in the ‘letterboxed’ version will compress very well which will save bits which the encoder can then use to improve image quality elsewhere in the frame. This is a tradeoff because it sacrifices the higher vertical resolution of the anamorphic format, but it won’t look too bad. In fact it may actually look BETTER on a 3:4 screen because the DVD player throws away the extra lines in the anamorphic format anyway when it ‘letterboxes’ the video for playback on a 3:4 screen (and it may do this without using the sophisticated re-sampling that the Virtual Dub or AVISYNTH filters can provide). Obviously with 2.35:1 'super-widescreen' films the effect is even greater.

So getting back to the topic, I prefer HC as an encoder overall, and recommend that you give it a try, but plan on doing some tweaking for best results regardless of which encoder you chose if you want really good performance at low bit rates.

[Mug Funky]

i'd vote quenc, but it's not on the list. however, all encoders have their quirks.

TMPGenc is alright, but it needs an RGB24 source, which means colours get molested, it is slow, and it seems to have "colour drift" where it moves to very slight pinks and greens rather than neutral grey on dark sources. this can cause "GOP blinking" if you have your lights off and your TV up nice and bright.

CCE is very nice (i voted for it because quenc isn't on the list) too.

Mainconcept could be better. it's nice and fast, but doesn't perform well quality-wise (even at maxed-out bitrate - 8.5 mbps CBR).

HC is a real contender, but has a couple of issues usability and quality wise (a GUI that keeps last settings would be awesome). also, i'm not sure if it's output is 100% DVD compliant. it'd be so cool if DVD was open-spec, so people didn't have to just guess at it based on the (open) mpeg-2 specs. however, the above could be said of all encoders, because when it comes down to it, not all DVD players interpret the specs in the same way.