For betatesters: TMPGenc 2.51: At last equal quality with CCE!

[SatStorm]

YES!
This version remains buggy, but for quality issuses I found it equal with CCE!

To test it yourself, I suggest to download both 2.51 and 2.51plus, and use the templates of 2.51plus on 2.51!
For best results, use this template: SuperVideoCD (PAL) - MPEG2 352x576 VBR 1020kbps.mcf
It creates CVD (the china's version of SVCD, compatible both with SVCD and Dvd's D2 resolution). It gives amazing quality, in really small size files. The quality is equal to CCE's 3Pass VBR!!!!

Now, I want beta testers to do a thing that it will help all of us a lot!
Please, you with DVD - R or DVD +RW recorders:
Encode a video with the SuperVideoCD (PAL) - MPEG2 352x576 VBR 1020kbps.mcf remplate and tmpgenc 2.51
Then, demux it, use maestro or whatever and burn a DVD with the file.
Inform the forum if it works!

Why we need this test when we have the compatible resolution?
Because DVDs has some special way for mpeg 2, which does not allow big gops. This new amazing template for 352 X 576, use big gops and then splits them on smaller ones. That's why Tsunami all the time say " Improve the gop, improve the split of big gops" etc on his history text the last months on TMPGenc.
Maybe, his technique creates amazing mpegs for CVD/SVCD, but the mpeg 2 produced file , which I remind you it is 352 X 576, a totally D2 LEGAL DVD resolution, ain't compatible with the DVD mpeg 2 standards! In this case, that amazing template is uselless...

If the mpeg 2 file is compatible with DVD and SVCD, then it is the total solution for mainstream use. SVCD for today and the same file for DVD tomorrow.
If the mpeg 2 file ain't full compatible with DVD, but only with CVD/SVCD, then it is useless for people wish in the near future to buy a DVD burner and made homemade DVDs without re-encoding their CVD/SVCDs

Hey beta testers, there is job for you now!!!

PS: This is my first post on this forum, so nice to be here!

[gldblade]

>The quality is equal to CCE's 3Pass VBR!!!!

Is there any proof of this claim?

Of course, I could try it out myself, but TMPGEnc takes an entire day (~25 hours) to work...

[ScooterMyth]

I really like this program but it's still TOO DAMN SLOW! Speed it up and CCE will find a good competitor. I encode in real time with CCE but with TMPGenc my encodes go to 3 or 4 times realtime. I keep checking each new version to see if it the encoding speed has been improved. One of these days it may (like when he starts asking money for it?).

[24hourloop]

You have to keep in mind that the CCE relevant core is coded in assembler. Considering the complexity of the process this is an enourmous feat. It'll be hard to beat.

Though I can't subscribe to your 3 to 4 times number. More like twice.

[chainsaw135]

LAugh@ScooterMyth

Well the fact is i dont do Pal SVcd..only NTSC so all PAL makers might have a great quality but what about us that have only NTSC 2 work with?

[SatStorm]

@ chainsaw135 : Well,you can use the NTSC template if you want, the results are the same.
@24hourloop: That you mention makes CCE faster. The quality issue is another thing. TMPGenc never had the quality of CCE also. Until the release of those templates (as part of the tmpgenc plus wizard). It use 2 pass VBR in a new way (there is also an option for the old 2 pass vbr...) and it use a very complex gop stracture...
The results are great.
@gldblade: 25 hours? Whow! Maybe it is a time for an upgrade? Sorry man, I show you something how to test it yourself, I can't do more!

@Anyone: You don't have to encode a whole movie! I need just someone to report that the mpeg 2 file generated by tmpgenc with those new templates with that gop, it is playable with a DVD-R or whatever. I use to have one for a while, but it wasn't mine, so I can't test myself! Come on people, one simply test!!!

Thanks!

[gldblade]

>Maybe it is a time for an upgrade?

It's fine for most things, excluding TMPGEnc. But the speed problem has already been mentioned...

[SatStorm]

Yeah, I use to say that for my old p2@233...
But now, with that encoding stuff I mess around, even xp+1700 ain't enough!
Arghhh, life use to be so easy without computers....

[da franksta]

dual athlon might do the trick for you ;)

[Dat]

I just tried using CCE with those resolutions above with better results then TMPG (in my opinion) looks like the battle still continues for me :D

Just gonna run some more tests now, maybe i will change my mind afterwards, but i doubt it.....

[ulfschack]

@Dat: What bitrate are you encoding in?

[Dat]

Video = CBR at 2498Kbps for both programs

[Atlantis]

It’s true that the quality has got equal to CCE. Only the new 2 pass VBR. I get excellent results. I always use TMPGEnc. CCE is a pain in the … It always gives different errors and I’m not a newbie! Never was able to make it work properly. And about speed, I get better results with TMPGEnc, with RGB files, that’s what I use. Yes CCE is faster with YUV files but if you have RGB files, TMPGEnc is better.

[ulfschack]

Hmmm ... my previous response was lost.

@dat: I was just wondering if you were using high bitrate or something, cause that's the only time I've seen CCE even come close to tmpgenc in quality. But I'm judging on the basis of how well the encoders treat my DV (home video) stuff, and that might be a different ball game than DVD-rippin.

Btw I use 2600 CBR, which doesn't leave much for the audio, which is crap anyways coming from a small microphone on the camera.

cheers

[SatStorm]

Now I am using tmpgenc 251 plus, "registered".
About speed, there is no improvement. The true is that tmpgenc 251 both standard and plus, have better support to SSE orders for athlon xp+ cpus... That is what improves the speed, and only a few...

About the quality: I am really impressed with the results, after hard testing. There are also some conclusions:
1. Don't use "soft block noise". It doesn't work as it used to be
2. Use the sharp filter between 74 - 127 both. That eliminates blocks for some strange reason!
3. Use gops like IBBPBBPBBPBB . Amazing results with 2PASS VBR (new type)
That is for low bitrate/resolution xSVCD. (let say 352 X 288 mpeg 2 interlaced, 1150kb/s)

For standard SVCD/CVD and typical bitrates (~2500), I don't see difference anymore between CCE and TMPGenc from DVD/DVB source. I still can see difference between them, when I use Avi source.... CCE remains better...

CQ_VBR of tmpgenc plus, also ain't the same as tmpeg standard. Split the gops better, so less blocks for us!

TMPGenc is much promising now. The days of CCE are limited!

[Dat]

Originally posted by SatStorm
Now I am using tmpgenc 251 plus, "registered".
About speed, there is no improvement. The true is that tmpgenc 251 both standard and plus, have better support to SSE orders for athlon xp+ cpus... That is what improves the speed, and only a few...

About the quality: I am really impressed with the results, after hard testing. There are also some conclusions:
1. Don't use "soft block noise". It doesn't work as it used to be
2. Use the sharp filter between 74 - 127 both. That eliminates blocks for some strange reason!
3. Use gops like IBBPBBPBBPBB . Amazing results with 2PASS VBR (new type)
That is for low bitrate/resolution xSVCD. (let say 352 X 288 mpeg 2 interlaced, 1150kb/s)

For standard SVCD/CVD and typical bitrates (~2500), I don't see difference anymore between CCE and TMPGenc from DVD/DVB source. I still can see difference between them, when I use Avi source.... CCE remains better...

CQ_VBR of tmpgenc plus, also ain't the same as tmpeg standard. Split the gops better, so less blocks for us!

TMPGenc is much promising now. The days of CCE are limited!

CCE's days are numbered you think? i totally disagree, but we will see in due time i guess ;)

[timekill]

I am making a SVCD of Pitch black right now using this template. It is bveing split into two files (for (SV)CD use for a friend, but I will try the first file as a DVD as well. If it works, I will attempt to merge the files, and then burn the complete movie as a DVD.

Note: I am encoding/demuxing/remuxing/joining etc on a PC, but burning to DVD on my mac - so this will be not only a dual format test, but dual OS as well =)

(Dual cpu too, as both my pc and mac are dual cpu: 2cpu.com rules)

BTW - I think it may also matter as much what DVD player you use as to whether or not this works. i.e. My Apex 5131 plays just about everything I've thrown in it, so just because it works for me (if it does) may not be a valid test...

I'll try to test it a number of standalones.

TimeKills

[Taxidermista]

Satstorm, I did your test. :)

TMPGEnc Plus 2.51.33.127
template SuperVideoCD (PAL) - MPEG2 352x576 VBR 1020kbps.mcf
Chapter 32 from Michael Mann's 'Heat'
Motion search precision: High quality
Play time 5 min. 40 sec. > encoding time 23 min. (Athlon XP 1800+)
Authoring with DVDMaestro, 4:3 aspect ratio (no problems here)
Added two ac3 soundtracks
Burned with Prassi PrimoDVD 2.0 (Pioneer DVR-A03, Pioneer DVD-RW media)

Looks fine (well...) in PowerDVD 4.0 XP. :)

Looks awful in my Sony DVP-S735D standalone but...it works! Smooth playback and correct aspect ratio.

I'll test it in a JVC player (picky) tomorrow.

[badbert]

Satstorm, in your reply you say:
2. Use the sharp filter between 74 - 127 both. That eliminates blocks for some strange reason!

can you please be more specific, are you talking about the sharpen edge filter? should both settings be set between 74-127? how do you decide which setting,between 74-127, to use? I am plagued by blocks in my lowbitrate encodings....any help would be appreciated...

[SatStorm]

@Taxidermista: THANKS!!!!!!!!!!!!!!!! I needed that Info!!!

Now, just imagine how many things we can play with!!!! ;)
Using big gop structure files (created with the new tmpgenc or even cinemacraft), burn them to DVD-Rs and became playable to very difficult players like Sony models, it is very - very promising!
The overall quality increase if you use big gops ;) (ok, that depands, but overall, increases....)
The template I mentioned, is a startpoint only. Now, I know that it works with DVD, so I can start testing to improve the quality. I believe that it is possible to create 4 - 5 hours (even more!) of D2/DVDs on 4.7 GB DVD-R discs, with near DVD quality! And that is great for mainstream use!!!

Thanks Taxidermista, you made my day!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

@ Badbert: Yes I talking about the sharpen edge filter.
I did many tests with that fuction this weekend.
My conclusions are:
If you have 720 X 576 (480) resolution from a DVD - video / DVB -s/-t/-c source, and you want to convert it to 352 X 576 (480), you can set this filter to values like 74, 84, 94, 104, 114, 127. That eliminates blocks... I believe 74 is enough, but sometimes higher values give better results
For DVB source with lower resolutions (like 544 X 576, etc), never set the value more 84...

If you convert 720 X 576 (480) to 352 X 288 (240) then there are more "tricks". For bitrates >1100, set the filter to 74. For lower bitrates (smallest possible: 800kb/s), set the filter about 94 and unmark the "Field Base" on the sharpen edge filter menu.
You get very fair results. Standard VCD like...

I use CQ_VBR for my tests, with 2 Pass VBR the results are better

I don't know if those parameters are good for analogue capture (.avi files). My tests are limited with DVD-Video and DVB transmissions!
Hope that helps someone ;)

@Dat: Well, CCE is a great and a very expencive proggy... I hope others, cheaper encoders became better CCE soon! I hate overpriced stuff... (even if I can find them on the net for free ;) )
TMPGenc plus made it and it is now equal in terms of quality with CCE(the new 2 pass vbr), but for speed it is waaaaayyy back.... Personally, I can do much more stuff with tmpgenc than CCE. For DVB transmissions, tmpgenc is the only way to go...

Again, Thanks ALL for your help!!!!!!!!

[DDogg]

Er, why didn't you ask people to do this test using 1pass CQ? I followed a thread on useNet that exhaustively discussed this and it seemed clear to me after reading all of it that the CQ mode (set 75-85) is clearly competitive to cce. Since tmpg can do it in one pass and cce takes 3 to 5 (including the VAF pass) the speed advantage is negated. Of course you lose file size predictability using this tmpg mode unless you do a couple of short runs and then extrapolate size results from these. That does make it more difficult but there seems no doubt the CQ mode does a damn fine job.

[Italbar]

DDogg,

Can you point me to the newsgroup that had the posts - or maybe post a summary here? I'd love to try and duplicate the effort and see what the "new" TMPG looks like. I love CCE, but frameserving with Avisynth and multiple filters with either 2.5 or 2.6 (even with VFAPI)is a hit-or-miss deal. Considering how much VFAPI and external filtering slows things down, I'm sure the encoding speed is pretty close (for 3-4 pass VBR at least.) If I can get the same quality from TMPG I am there all the way.

Italbar

[Okar Gona]

I have tested 2.51 on some of my DV recordings and Analog Video. The results are very good. It is at least equal to 5 pass CCE with average bitrate of 2300.

I have changed the default SVCD settings a bit:
Audio 160 kbits
Max Video 2584 kbits
Motion search: Highest Quality
CQ_VBR = 100 (Size of output file is not important for me)

Depending on the video source (high motion or not) size maybe greater or smaller than a MPG-file created with CCE.

Yes it slower than CCE but for video recordings of 30 minutes or so I don't find this a problem.

Together with TSCV you have got two easy to use tools to make the perfect SVCD for your own recordings.

[Arky]

What of tweaking the TMPGEnc matricies?? MB1 seems to be the man who knows about this stuff...

It would be really cool if we could all develop a template, and post it to a Yahoo Briefcase for everyone, or even have Doom9 host it (after all, a template is only a couple of kb!)


Arky ;o)

[DDogg]

Italbar,

It is just the method described in "Dr. Mercury's Movie Kit" always available in Alt.binaries.multimedia.utilities

It does produce an *excellent* encode but is rather labor intensive. His point is the TMPG CQ mode is where it is at and he has a large following. There was some long rambling posts making points for both TMPG and CCE in A.b.vcd. SOunds like they can be about equal. The kit is worth a download if only just for another person's perspective.

You got to take all this stuff with a grain of salt until you do your own encodes and judge them yourself. There are so many "pundits" out there...half of them have are full of opinion coming from a half assed brain. That said, this kit and guide looks pretty good, and as I said, I think it is worth the 10 meg download.

If I were going to use TMPG for my DVD encodes I would definately do it the way described in this guide using CQ mode. I will use TMPG for my DVCam stuff for sure.

The latest DVD2SVCD alpha has limited support for TMPG CBR and 2pass VBR. Maybe one day it might support CQ mode. If any of you have time, please test some chapters using dvd2svcd in both CCE mode and then the same in TMPG mode.

[phuntyme]

I've been playing with the CQ mode since someone pointed me to a web page that described a method to fit a 90 min movie onto a SVCD. To my surprise, the quality was superb and I haven't turned back since!!

That bit rate calculation thing is unfortunate. It would be nice if we could, not I noticed that there is a log file, and it includes the bit rates of different frames. Can I use this for anything? I'm sure I can sort it out into an excel spread sheet to be able to recalculate things... Its just I prefer to do a maximum of 2 encodes to get the right size rather than 3 or more, it becomes kind of a drag when this happens.

[SatStorm]

The reason I ask to test that special template, is really simply.
That wasn't a test for quality, it was for compatibility. That is what the people on all forums don't understand.

Quality is another issue.

Personally, I use only CQ_VBR for my encodings. The quality is amazing. I see no blocks on movies. There are some blocks when I encode videoclips from music channels, with extreme action. Using CQ_VBR 65% with max 1350, for 352 X 288 gives amazing quality (vhs like)
With the new tmpgenc plus, the speed is about 0.50X for any souce to 352 X 288 mpeg 2, and 0.30X for any source to 352 X 576 with my athlon xp 1700+

Just for the info, xSVCD with low resolution/bitrate and the use of tmpgenc for the encoding, it is possible to create VHS like media, up to 95 min. Using extreme technics, you can easily go beyond 100 - 110 minites.. The more "extreme" you go, the less compatible you are.

Those things has nothing to do with the quality. But try to remember, that there are 2 kinds of users: Those who wants quality and those who want the more possible on a CD...

I am a mainstream type, I want both :D
That is the reason I want to learn how to create the best CVD possible, just to be prepared for D2 DVD creation. I strongly believe that D2 DVD (352 X 576/480) is for the future, what VHS is today. And I have no problem with that!

People who wants quality, maybe it is wise not to read those post. They getting ungry for a strange reason. They can't understand that for some people, Fair Quality is enough!

@phuntyme: There is a "trick" for CQ_VBR and 352 X 288 xSVCD. Go to any bitrate canculator. Go to svcd settings and set CD capacity, movie Lengh and the audio you want to keep for your VBR encoding.
On the result add 100. For examply, if the calculator say: 1000, you have 1100. Then go to tmpgenc,and on CQ_VBR set higher to that value, 1100 and the quality bar to 75%. If the movie you encode has extreme action, maybe is wise to set it to 65% or even 50%!
Always de-interlace the source and set the encoder to create with the rest 288 (240)vertical lines interlace output! Use the sharpen edge filter about 74 both and don't use "soften block noise"
With those settings and the bitrate trick, most of the times you end up with 780-800MB file. Any 80min CD can hold that files (overburning allow even 810MB files)

hope to help someone!

[Italbar]

Ddogg,

Thanks! :)

I was just glossing over those posts on Usenet. It seemed to me that it was a flamewar between a DVD2SVCD poster and Dr. Mercury's followers. I have never looked much into his guide I have to admit....I downloaded it once, read up until I got to the normalizing audio portion, where he states of the program:

"It'll 'normalize' it 100%, or double the volume".

As I was always under the assumption that normalizing just found the loudest portion of the sample and increased the volume of the entire sample to make that point 100%, the guide lost credibility for me at that point and I never read further.

I'll check it out if it has a higher quality encode method.

I assume the labor intensive part of the guide is due to the unpredictability of the CQ file size?

Thanks again.

Italbar

[DDogg]

"I assume the labor intensive part of the guide is due to the unpredictability of the CQ file size? "

Yeah, that and the audio method.

Italbar, it would be great if you (or someone) had the time to play with the different methods for audio. I don't know HOW we can get much better than the BeSweet routines for downmixed audio. I don't think I have the level of equipment to tell :(

Sounds like the usenet guys may be behind the times on this subject, but we always have to keep an open mind or we stop learning, yes? Anyway, if we were to think this older method better we could mention it to dspguru. That would make him crazy and he would have to invent something better for us :)

DVD2SVCD mentioned in the sticky thread about TMPG in the advanced section that was working to come up for a credible plan to attempt file size projection for CQ mode. That would be some real nice icing on the cake for sure.

[Arky]

Hmmm.... I just tried the template in question qith DV footage, and I agree that it gives good results, but I do not feel that it is really that much better than previous results with older settings.

Furthermore, I have succeeded in importing the files into both Maestro and Scenarist, #~BUT~# Scenarist rejects the files unless you have set "Max GOP length" to 15 (PAL), which is hardly surprising, considering the attempted violation of the DVD spec. I have not established yet, whether TMPGEnc manages to create "Quasi-GOPs" which pretend to be no longer then 15 frames long, but nonetheless succeed in making greater use of P and B frames, by some clever techinique.

So, promising results, but nothing nearly as revolutionary as everyone else seems to think... :|


Arky ;o)

[SatStorm]

It is a start.

"One Small step for quality, One great step for Encoding..."
SatStorm 13 - 02 - 2002

:D :D :D :D :D :D :D :D :D :D :D :D :D :D :D :D :D