Who wants to test a new MPEG2 encoder

[doxville]

Originally posted by zettai
Everytime I try to load an avs file into HC it just hangs. No crash error message, just no response. The avs file works fine in vdub.

I have an old fashioed althon xp, using win2k. Suggestions? Same with me, I am also using Win 2k and right after opening the source-avs, the programme hangs up with no response, no matter how simple the script is (colorbars, etc.). I am using the latest Avisynth (beta2), but that shouldn't be the problem, right? Both Versions of HC, 0.1 and 0.11 have the same problem... :(

[hank315]

Just installed the latest Avisynth beta but everything runs ok.
Using XP, SP1. Don't have a machine with 2000 to test it :(

[danpos]

Originally posted by hank315
Just installed the latest Avisynth beta but everything runs ok.
Using XP, SP1. Don't have a machine with 2000 to test it :(

Hi, hank! You could to try a win2k emulator like VMWare, QEmu, etc ... In this way, you can to test HC in win2k environment without the necessity of do partition on disk, to install the system, blabla ... It's just a suggestion!;)

Keep up the good work,

[hank315]

Will u add some kind of SMP support in the encoder pipeline? and yuv2 (4:2:2) support?SMP support is one of the things still to do but it is not easy to synchronize threads in an encoder, already tried it...

4:2:2 colorspace: probably not until I really need it.

Concerning HD res, there is alot more to it besides being able to include different profiles? Obviously I have no idea, just curious.Also don't know much about it but what I've seen so far Main profile @ High level could be implemented.
BTW, what exactly are the specs for HD?

[LigH]

On my AMD Duron 800, HC 0.11 does not detect that this processor supports MMX/ISSE, and creates a file "fort.24" with the content
No MMX/SSE detected
HC needs atleast MMX/SSE to run
aborting...
<ENTER> to close program

But afterwards, HC 0.11 GUI starts, and works well. (BTW: Great quality!)

[Amnon82]

I think it is GUI related. If there is a DLL version of the HC Encoder my DelphiGUI will work ;) Also the commandline version works with my gui.

:D

... or let us know which DLLs You're using for Your GUI ...

[hank315]

@LigH
That's strange because this message was removed, are you sure it's not an old message from the first version?
If MMX is present and SSE not, it should generate a warning but I couldn't test it, don't have a Duron or old Athlon around here...

[hank315]

... or let us know which DLLs You're using for Your GUI ...????
Just the normal DLL's are linked when you use API calls...
But if you want to know, it imports stuff from kernel32, user32, gdi32, comdlg32, comctl32, ole32 and oleaut32, so nothing special...

[zettai]

Solved my particular problem.

I was only quickly opening it up to have a look at it so instead of unpacking it, I ran the .exe by double clicking in winrar (which should just extract to a temp foler and run the executable). This was the version that would fail to open any scripts - if I extracted to a regular directory the normal way it would be fine.

No idea why, but there you have it.

[robot1]

Originally posted by danpos
Hi, hank! You could to try a win2k emulator like VMWare, QEmu, etc ... In this way, you can to test HC in win2k environment without the necessity of do partition on disk, to install the system, blabla ... HC works fine in win2k

[JetlagMk2]

Originally posted by robot1
HC works fine in win2k
Works fine on my w2k sp4. Thanks, hank!

[doxville]

Works fine on my W2k SP4, too... Now...;)
I found out that antivir (free-av.com) causes all the problems. After having deactivated the "AV-guard" the encoder opened my *.avs without problems, giving me opportunity for heavy testing! :D
Thanx, hank!

[rhayman]

First off i really like this encoder and think it has great potential.

Secondly i notice the program has a lot of built in martix's, i have been interested for a while to find out about these. However im useless at searching forums and couldent really find anything specific enough to help me.

My question is, what is the difference between the matix's, are some better quality than others, when should you use them etc. But mainly do they all create standard mpeg2 files for dvd authoring.

Sorry if this has been explained before and i've just missed.

Any help is appreciated.

[LigH]

Originally posted by hank315
@LigH
That's strange because this message was removed, are you sure it's not an old message from the first version?
Uh - probably; I tried all versions around the same time.

AMD Duron supports MMX and ISSE. I just wonder why it didn't detect it.
__

@ rhayman:

Different matrices are useful for different purposes.

For example, there are matrices which have the potential to encode the video very exactly - but at the cost of high bitrates. You will notice the rather low values even near the right bottom area of the matrices.

Other matrices will allow a quite fair quality even at low bitrates. But they will reduce details and provide a rather blurry result. You will notice quite high values especially at the right and bottom borders.

[Amnon82]

@rhayman: I and others explained it in this thread already:
http://forum.doom9.org/showthread.php?s=&threadid=89897

take a look ;):search:

[rhayman]

Sorry, i did a search before that was posted and could not find anything, i should of searched again before posting.

Thanks for the link :)

[Amnon82]

no problem

[hank315]

Seems that people who are running anti-virus programs have trouble running HC, it could take a long time to start because the exe file is packed with UPX, some anti-virus programs don't like that.

So I changed the download version to a non-packed executable.
Just download it again and see if it helps...

Still the same link: http://hank315.dyndns.org/HC_011.zip

[gigaset2004]

@ hank

I have tested your encoder with bitrates at 1000 with several files and i also tested bitrates at 2000 and rejig them.

I found the encoder pulses the B frames ?

The quality is not so nice at both ways. I know you said the encoder is normaly for DVD ranges.

Maybe it is possible to explain what happens wit the B / P Frames ?

Nice work anyway

[doxville]

Originally posted by hank315
Seems that people who are running anti-virus programs have trouble running HC, it could take a long time to start because the exe file is packed with UPX, some anti-virus programs don't like that.

So I changed the download version to a non-packed executable.
Just download it again and see if it helps...

Still the same link: http://hank315.dyndns.org/HC_011.zip
Sorry, but I'm afraid it doesn't help with my antivir-software. I still got the same problems as before with the unpacked version... :(

[Xeno86]

Hi everyone!

@ hank315 & dragongodz
I didn't have time recently, but I think that I've found why the quant doesn't go below 5.3. The problem is connected with mild noise. Anyway this is my (new :) ) test script, so you can duplicate my results:
BlankClip(length=1500,width=528,height=576,pixel_type="YV12",fps=25)
Blockbuster(method="noise",variance=1.5,cache=8192,detail_max=64,seed=5623
Encode it with default settings (using public ver. 0.11beta) with:
1) bitrate=2000 and max=2500
2) bitrate=2000 and max=6000
3) bitrate=2000 and max=8000

What you can see in bitrate viewer is that 1) and 2) looks almost exactly the same - only first 6-7 secs differ. The rest has quant=5.3 and bitrate=550-650. So it is not low max that spoils the result ;)
3) Looks totally different but very strange though - bitrate jump from 950 to 4000 kbps.
A quick test of my initial clip without this BlockBuster line has shown that bitrate raised. I'm doing a full encode now though. I'll post back later. It's strange that bitrate goes lower with this line, because it add more 'details' (noise) which should be more difficult to encode and bitrate should go higher, shouldn't it?

Btw. I like the new version :D

[hank315]

Posted by FreelockStill needs a temporal filter to correct grain (convolution3D).
Deblocking is a good idea too (thanks Dragongodz).
So, for the next release, can you give us, Hank, some advanced settings to adjust?
Is it possible to do that with HC?No, not ATM. It just does a straight forward encoding without filters etc.
Ofcourse it can be programmed but it will certainly not be in the next version.

@Gigaset2004
Bitrate is optimized for DVD creation, average bitrate > 2500-3000.
Pulsating B frames just means the nr. of bits for these frames is too low. In the next version I will try to get a better distribution between I, P and B frames at low bitrates.
But if you're doing 1000 Kb/s at full DVD resolution results will always be bad, IMHO the bitrate is just too low for MPEG2.

@Xeno86
Will try that noise stuff when I can find the time.
There's so much to do and so little time ;)

[dragongodz]

Xeno86 - i have had a look at that test. hmm interesting. i have emailed hank315 about what i see etc. so we will just have to wait and see what he makes of it aswell. :)

[freelock7]

HANK315-No matter. I found these settings for each encoder to produce the same great quality (VBR=2Pass/2Pass):

CCE2.70
cpu=4
Low pass filter: natural3* (32/10)-by default
Levels(0, 1, 255, 3, 255)
undot()
lanczos4resize(720,576)
converttoyuy2(interlaced=true/false)
-----------
QuEnc0.59B2
cpu=4
undot()
lanczos4resize(720,576)
-----------
HC0.11
cpu=4
Convolution3d (preset="movieHQ")
Blockbuster(method="noise",detail_min=1,detail_max=3,variance=0.1,seed=1)
undot()
lanczos4resize(720,576)

[Xeno86]

A quick test of my initial clip without this BlockBuster line has shown that bitrate raised.
Here are the final results of full encode:
All settings as before - best profile, notch matrix, gop 24 2, length 0:50:35, 528x480, 23.976 fps, max.bitrate=2500, bitrate=2000.
avg.bitrate=1691 (computed value; bv shows 1584, hc shows 1722, mpeg stream eye shows 1691)
avg.quant=4.16
min.quant=2.34
Filesize=637415901 BPrevious results (same source & settings but with Blockbuster(method="noise",variance=1.5,cache=8192,detail_max=64,seed=5623) ):
avg.bitrate=1247
avg.quant=5.54
min.quant~5.0
Filesize=469897135 BAs you can see there is a huge difference.

[dragongodz]

Xeno86 - dont get too caught up by quants rising. for harder to compress ,but bitrate limited, encodes i would expect to see the average bitrate rise noticably and the minimum bitrate to rise maybe slightly. obviously the minimum bitrate is rising more than a little and is a problem but we will have to wait for hank315 to have a look and see what he can do in code.

[hank315]

@Xeno86 and dragongodz
Just started some tests using noise with Blockbuster.
Using:
BlankClip(length=1500,width=528,height=576,pixel_type="YV12",fps=25)
Blockbuster(method="noise",variance=1.5,cache=8192,detail_max=64,seed=5623)gives bad results but the variance is just too small, the "energy density" of the planes is low, a small difference in Q will give a huge difference in bitrate.
Example: Q=2 --> all coefficients will be coded, Q=3 --> almost all coefficients will be zero.

Doing something likeBlankClip(length=1500,width=528,height=576,pixel_type="YV12",fps=25)
Blockbuster(method="noise",mean=80.,variance=200.,cache=8192,detail_max=64,seed=5623)would be better for testing but even then the bitrate should be better.

So there's still some work to do ;)

[fewtch]

I just tried this encoder for the first time. I'm a newbie to MPEG2 encoding, so bear with me...

First off I chose an average bitrate of 3700 and max of 7000 in order to fit on a DVDR along with some chosen audio. I used a D2V file (DGMPGDec v1.2.1) indexing the movie "Carlito's Way" from decrypted VOBs (NTSC, pure FILM) and changed aspect ratio to 1:1 in HC. I didn't change any other defaults except that and bitrate.

HC produced an M2V that TMPGEnc DVD Author rejected as being out of DVD spec (23.976 FPS instead of 29.97). This part I don't understand, but it's probably something very simple I missed... this is where I'm new to this sort of encoding (no big deal, it was just a test). Also, there were way too many artifacts in the final output for my taste, but clearly the bitrate is on the low side. The film itself is quite badly grainy as well and not very good quality for a DVD movie.

Not sure about the aspect ratio thing, I think the movie itself is 2.35:1 so dunno which I should have chosen. 1:1 seemed to make the most sense for re-encoding from VOBs, someone correct me if I'm wrong.

Anyway, just some output from a newbie encoder's experience with the program, hope it helps. Version is HC 0.11b with GUI.

[dragongodz]

out of DVD spec (23.976 FPS instead of 29.97). This part I don't understand, but it's probably something very simple I missed.
very simple indeed, 23.97fps is not a valid DVD framerate. to use 23.97fps footage you have to use pulldown which of course is equivilent to making a fake 29.97fps. lookup telecining for more information about that. ;)

Also, there were way too many artifacts in the final output for my taste, but clearly the bitrate is on the low side.
as you can see from the last few posts there is still work to be done on the rate control.

The film itself is quite grainy as well
and of course there are avisynth denoisers that can help clean up the source video if its a bit grainy to start with, which should look better if not over done.

Not sure about the aspect ratio thing, I think the movie itself is 2.35:1 so dunno which I should have chosen.
personally i have only ever seen 4:3 and 16:9 on dvds. even when they say another ratio they have had black bars added to make them either of those. 1:1 is definatly not right. :)

[fewtch]

Originally posted by dragongodz
very simple indeed, 23.97fps is not a valid DVD framerate. to use 23.97fps footage you have to use pulldown which of course is equivilent to making a fake 29.97fps. lookup telecining for more information about that. ;)
Isn't the majority of FILM material out there at 23.976fps (at least with NTSC)? If so, how about a warning from HC that the result won't be DVD compatible, since it seems to me re-encoding from VOBs to get something that would fit on DVD-5 wouldn't be an uncommon usage...

personally i have only ever seen 4:3 and 16:9 on dvds. even when they say another ratio they have had black bars added to make them either of those. 1:1 is definatly not right. :)
I'm still really lost when it comes to aspect ratios and such. When I used 1:1, black bars were there in the resulting .m2v... are you saying they would not if I used 16:9? Remember, this encoder has an option to use .d2v files, which if I'm not mistaken are always derived from DVD's.

P.S. sorry for any ignorance I'm displaying here. :confused: My experience is mostly with DVD->AVI and DVD->DVDShrink->DVD. Working directly with MPEG2 is still brand new stuff to me. Just thought a newcomer's perspective might add something to the discussion (then again, maybe not :rolleyes: :D).

[FredThompson]

Accounting for pulldown is a good idea. For the rest, you really should do some research on the basics.

[LigH]

MPEG-2 allows 4 different aspect ratios, in general. There are:

- 1:1 Pixel Aspect Ratio [= PAR] (not allowed on DVD!): The content shall not be stretched in any direction upon playback.

+ 4:3 Source Aspect Ratio [= SAR] (valid for DVD): The content shall be stretched slightly, so that the resulting image will have an aspect ratio of 4:3 (this would stretch lines with a width of 704 pixels to a width of 768 lines; if 720 pixels shall stretch to 768 lines width, too - or to rather 786 pixels instead, is a topic for philosophers for ages now...)

+ 16:9 Source aspect ratio (valid for DVD): Similar topic here, the content shall be stretched remarkably, so that the resulting image will have an aspect ratio of 16:9 (means, 704 pixels width to 1024, 720 to probably even more). This is known as the usual "anamorphic" encoding.

- 2.21:1 Source aspect ratio (not allowed on DVD!): Stretches even wider upon playback...
__

So what if your movie has "the interesting part of the image" with Cinemascope aspect ratio? Well, DVD allows only 4:3 SAR (704:768 PAR) or 16:9 SAR (704:1024 PAR). If the interesting part of the content would have a SAR of 2.35:1 (as often printed on the back side of the DVD case), it still can only be encoded at an anamorphic aspect ratio of 16:9 SAR. So it still needs some small black bars ("letterboxes").
__

And about d2v project files: They store the positions of the GOPs (or I frames) inside an MPEG file. But they don't store the aspect ratio - the safest way to determine the aspect ratio of a DVD, is to look at it and decide! DVD2AVI / DGIndex is only able to report bitstream flags. And it is possible that these are wrong (if the encoding studio had set up the encoder incorrectly - and this happens from time to time...).

Keep in mind, that DVD video may only have Source Aspect Ratios of "4:3" or "16:9". Use the thumb rule: If the actors have "egg heads" in the preview of DGIndex (press [F5] after loading the VOBs, and later [Esc]), then the SAR is "16:9" (well - except for the movie "The Coneheads", maybe); if their proportions look quite fine, it is "4:3". This is the best way to decide - software is not able to do so.

[fewtch]

If I recall correctly, with this DVD the view in DGIndex showed the video *squished* vertically (not stretched), i.e. heads too much rounded rather than egg-shaped, and with large horizontal black bars already there in the DGIndex display. If it matters, this is an old/OOP DVD from 1998. The back of the DVD does say "2.35:1 widescreen." Another weird thing is that the main movie, with no extras and few audio/subtitles takes up over 7.3GB on the disc. I'm thinking this may be an old, nonstandard encoding of some sort.

Anyway, thanks... that helps some, although some confusion remains (mainly on the framerate issue -- why pulling video directly from the VOBs using .d2v gave a DVD-incompatible framerate with HC?).

Will do some research, or if you know the specific answer maybe PM me (if you have time)... I'm done.

[Orion|69]

@ hank315 : respect.. i truly love your encoder already, altho it's ofcoz not finished :)

what i prolly love most is that all of this is done in fortran and an additional pure asm lib :)

since the company i worked for ditched fortran in the, lemme think, early nineties (i think around 93/94), i never saw such a great tool produced with the language :)

u made an old coder smile :D

respect also goes to the ongoing efforts of the betatesters and the input of the not-so-betatesters :) i'll try to do my share from now on :)

[dragongodz]

why pulling video directly from the VOBs using .d2v gave a DVD-incompatible framerate with HC?
thats what you havent understood. its NOT incompatible its just not a legal framerate used as is, so it has to be telecined(2:3 pulldown).

now different authoring programs will handle these diferent. from some checking i just did Tmpgenc it appears will just error and not accept it without pulldown already done. while DVD-Lab it appears will give you a warning and do the pulldown itself.

so yes maybe adding an option to output an already pulldown set stream is a good idea.

some reading on how telecining works can be found here
http://www.doom9.org/index.html?/synch.htm

The back of the DVD does say "2.35:1 widescreen.
and if you were to cut off the black bars thats what it no doubt would be. they still do that on dvds, quote the picture part(minus bars) aspect ratio. with the bars though you will find its either 4:3 or 16:9.

for those interested in dvd aspect ratios heres a little read aswell
http://www.doom9.org/index.html?/aspectratios.htm


anymore general dvd questions should be in a new post please.

[JetlagMk2]

hank, could the GOP structure be manually set? Maybe from right out of a d2v file? I'm sure you already have a lot to do it's just a humble feature request. Thanks for the great program!

[Trahald]

Did the .ini support go away in the gui versions.. i dont see anything specifically mentioning it but have not been able to get them to read the .ini

[hank315]

Did the .ini support go away in the gui versions.. i dont see anything specifically mentioning it but have not been able to get them to read the .iniYes it did...

But ATM i'm working on a batch version which will take command line arguments in the way Quenc does.
It will run in a command prompt window but the encoding engine is the same as the GUI version.
Not every option will be supported, some general options will be in the ini file but the next arguments will be supported:
-i ---> input file
-o ---> output file
-b ---> average bitrate
-maxbitrate
-aspectratio
-ini ---> HC.ini file for all other settings

Doing it this way it can also be used with DVD-Rebuilder.
Will be ready in a week or so...

[manolito]

Originally posted by hank315

Not every option will be supported, some general options will be in the ini file but the next arguments will be supported:
-i ---> input file
-o ---> output file
-b ---> average bitrate
-maxbitrate
-aspectratio
-ini ---> HC.ini file for all other settings


I think you should also support the "Interlaced" option as a command line argument. It is not important for DVD2SVCD, but DVD Rebuilder needs it because quite often some of the extras are interlaced while the main movie is progressive. Ask jdobbs...

Cheers
manolito

[FredThompson]

Originally posted by manolito
I think you should also support the "Interlaced" option as a command line argument. It is not important for DVD2SVCD...Huh?!?! The ONLY source DVD2SVCD accepts is progressive?!?! Specifying interlaced source most certainly is important.

[manolito]

@FredThompson

No no no, big misunderstanding! Of course you have to be able to specify interlaced or progressive when using any encoder with DVD2SVCD. It is just not absolutely necessary to support "Interlaced" as a command line option. In the case of HC there are many HC settings that are unique and do not have a corresponding option in QuEnc. So it will be necessary to also have a template for HC where these settings can be made independently of the command line options. Since DVD2SVCD only encodes the main movie, you can set up "Interlaced" or "Progressive" in the template.

For DVD Rebuilder it is a different story. It encodes the whole DVD, and it happens all the time that there are interlaced and progressive parts on one DVD. So Rebuilder must have a way to tell the encoder how to treat every part of this DVD independently.

Cheers
manolito

[Trahald]

@hank
excellent! being able to set tff/bff would be nice too (via ini is fine.) also start time of 00:00:00:00 as an option (i dunno if the gui versions changed since im still using .01)

[Mug Funky]

one thing that would be pretty cool is a "save settings" option in the GUI. possibly saving it to the old-style .ini file. i have to re-set everything every time i open the program.

i eagerly look forward to batch mode though :)

really great work on this encoder.

[FredThompson]

The graphic version hangs when I feed it this script with the intent of making an MPEG2 still:

ImageSource ("custommenu.bmp")
ConvertToYV12

Is the current release incapable of making stills?

[hank315]

Is the current release incapable of making stills?Just did a test with your bitmap, result at:
http://hank315.dyndns.org/stillmenu.m2v
one I frame, size 180 kB, Quant=1
One frame is a little bit difficult to play, loaded it in cuttermaran, looked good.
Turned the I frame into a bitmap again:
http://hank315.dyndns.org/still.bmp

[FredThompson]

Wow! That looks great! How did you make it? The GUI version freezes when I tried.

MPC plays the still just fine, btw.

[hank315]

Used the same avs file as you did, the first time got a "movie not progressive" error, second time was OK.
Think I'm going to dump this check...
Using the last Avisynth beta release.
Start HC_011 and change the next default values:

endframe 1
dc precision 8
activate Constant Q: value 1
aspect 4:3
matrix QLB

This should generate the same m2v file.

[FredThompson]

I've tried to set HC.ini with your settings but I'm seeing an error.

*INFILE custommenu.avs
*OUTFILE menu.m2v
*PROFILE BEST
*ASPECT 3:4
*CQ 1
*DC_PREC 8
*FRAMES 1 1
*MATRIX QLBcustommenu.avs, avisynth_c.dll and the entire HC set are in C:\Working\DTiVo\

If I launch HC.exe, AviSynth reports custommenu.avs cannot be found.

With this setting:

*INFILE C:\Working\DTiVo\custommenu.avs there is another error showing there is an attempt to read from:

C:\Working\DTiVo\ C:\WORKING\DTIVO\CUSTOMMENU.AVSWhy would this happen? Is it a parsing error in HC?

[hank315]

Used the GUI version to create it, and the avs file opened the second time I tried it :confused: think the command line version will just abort on that one.

Please try it again with the GUI version, that might help.
Will remove some avs checks in the next version that don't seem to work OK.

[FredThompson]

Ah, the command line version is a little behind the GUI version, huh? You'd almost think this was a one-man hobby project :)

The error is an AviSynth error, not HC. It looks like HC is screwing up the path when it calls AviSynth. I've tried the GUI version and it keeps hanging. Maybe it's got the same issue with opening the AVS file as the command line version. The installed AviSynth here is AviSynth_210205 and avisynth_c.dll from avisynth_c-0.15. Which are you using?

I was actually hoping for the command-line version. The root of this is mpeg2enc problems with TyTool. HC looks like a far better option and it has the command line interface which is a far easier to control than GUI automation, though it can be done. TyTool allows removing sections within an existing MPEG2 stream and only encodes the modified GOPs. Part of this process is calling mpeg2enc for each "new" GOP. That, and the single frames used for menus, are the parts which could benefit from HC. mpeg2enc is also lousy for HD content.

This is very impressive output, by the way, VERY impressive. I've got another "torture" background which looks like a red and green Christmas package decorated with gold bells. It makes just about everything else puke, especially if the menu text is white. Most encoders hate the high contrast of saturated red and green near each other and they smooth out the green too much. Here's that background image: http://home.mindspring.com/~fredthompson/Christmas Gift Wrap Title Page.zip The yellow edge of the ribbon was added to compensate for the corruption of most encoders. It could be replaced with the same red as the rest of the ribbon for an even harder test. The texture in the green was added in the hope it would reduce noise around the white menu text. It could be replaced with a uniform green without any objections. With the texture, it would provide an interesting test.