Newbie stuck @ mpeg2 > xvid conversion

[kamaleon]

Hi folks

I'm really sorry to have to come up for help but I'm really stuck here. I'm trying to encode most of my PAL DVB-S recordings to xvid to gain some place (hard drives are expensive )

Basically these are .mpg that've been through VideoReDo Plus and I have TMPGEnc Xpress.

But there's two things I can't figure out:

If I encode at 720x576 pixels (the original resolution) the encoding appears stretched, should I encode at 768x576? I've never understood the thing about "true" 4:3 display ratios. Basically all I want is to watch my videos on my computer.

Another thing, what about fps? should I encode @ 50 fps (progressive)? When I choose 25 fps videos seem choppy, not fluid, like I have the impression of watching an ancient BW silent film the ones in which you can see the frames

Thanks a lot for all your help.

[Baldrick]

Convert to square pixels(1:1) with a 4:3 resolution like 720x540, 640x480. If your source is 16:9 then use 16:9 resolutions like 720x404, 640x360. Don't upresize to 768x576.

Convert the mpeg2 with AutoGK and you just have to set file size and resolution and it will fix everything else automatic. .

[kamaleon]

Hi Baldrick, thanks for the quick reply.

I've used AutoGK once but it had some spyware?

I'm using TMPGEnc and it seems to suit me, but thanks anyway.

Is there any option for 1:1 on TMPGEnc?

So isn't PAL 720:576 1:1 ? I really don't understand this.

[Baldrick]

Latest version of autogk doesn't contain any spyware.

But tmpgenc xpress should work fine too. Can't you set a custom resolution/frame size?

[kamaleon]

Latest version of autogk doesn't contain any spyware.

Ok, thanks, good to know.

But tmpgenc xpress should work fine too. Can't you set a custom resolution/frame size?

Yes, I really enjoy it so I'd rather stick with it.
Thing is, I'd really like to understand the resolution thing... i thought 720x576 was 4:3?

also, what about 25 vs 50 fps?

[Baldrick]

The dvb 720x576 video has a display aspect ratio(DAR) flag with 4:3 or 16:9, so when you are playing it will be resized to 4:3 or 16:9. You can also convert to 720x576 avi xvid with 4:3 or 16:9 DAR but it wont work in some video players that ignores the DAR in avi files. That is why I convert to a resolution that is square pixels, it has same DAR as the resolution. Confusing... read instead http://www.doom9.org/index.html?/aspectratios.htm

I leave the framerate issues to someone else....

[kamaleon]

Read it, still it's very confusing, yes.

But if 720x576 isn't 4:3 then what is it?

If I resize it to 768x576 what am I making wrong?

If I'm going to destroy the original files (dvb-s recordings as well as DV captures from camcorder) then I wanna make sure I won't do anything wrong...

[LOWTECH]

Where do you get the idea that Hard Drives are expensive, I have just bought a 500GB external Hard Drive for £65 in the UK at Comet

Very cheap now, and would save you all the bother and retain the quality.

[kamaleon]

Everything's relative mate, I'm on a low income...

[mats.hogberg]

Most standalones don't play AVI (or anything else for that matter) with a resolution > 720 wide. And you'll just lose quality by resizing up.
In PAL land, "all" DVD video is 720x576. This can be displayed at either 4:3 or 16:9. It can also be 352x576, and still be 4:3 or 16:9.
With AVI video, there's no reliable way to set the DAR, so you'd better make the AVI resolution the correct DAR, and this is best done by adjusting the vertical resolution to match your desired DAR.
(BTW, 720x576=1.25=5:4)
If I was you, I'd just use AutoGK, and let it figure out all the details for you.

/Mats

[Midzuki]

4:3 720x576 should be resized to 1:1 704x528.

------------------------------------------------------

Midzuki would have written:

But if 720x576 isn't 4:3 then what is it?

720x576 is 4:3 IF it has a 4:3 AR flag.

720x576 is 16:9 IF it has a 16:9 AR flag.

720x576 is 5:4 IF it has a 1:1 AR flag.

If I resize it to 768x576 what am I making wrong?

Quite probably your standalone player does not support it.

[kamaleon]

If I was you, I'd just use AutoGK, and let it figure out all the details for you.

How good is AutoGK's deinterlacing compared to TMPGEnc's?

If I resize it to 768x576 what am I making wrong?

Quite probably your standalone player does not support it.

That's not a problem, I don't have a standalone player, I only ever watch stuff on my computer. Is there anything else wrong with it in that case? I understand upscaling/resizing to 768x576 isn't probably very good, but what about downsizing to 704x528 or 352x576? Is that any better?

Which brings to another question: not all DVB-S transmissions I've recorded are actually 720x576, they have different resolutions. I'd have to have a look to see exactly how many different ones and post them here later. What to do with all of those different resolutions?

[manono]

How good is AutoGK's deinterlacing compared to TMPGEnc's?

Probably better. While you're doing all of this writing and asking of questions, you could have been running some practice encodes. Reencode some videos. Try out AutoGK and/or TMPGEnc for the job and decide for yourself. One thing, though; AutoGK will make all the hard decisions for you, and make them right. You on the other hand, if using TMPGEnc together with your apparent lack of knowledge, are more than likely to make some wrong ones.

[mats.hogberg]

What to do with all of those different resolutions?

Just select a resolution that retains the DAR - 4:3 or 16:9. AutoGK will do that for you. About deinterlacing:

KernelDeInt filter is a filter performing deinterlacing in AutoGK. Its threshold parameter allows you to control sensitivity of the filter in detection of motion areas (as static areas are ignored by the filter). This is very advanced settings and it should be used only when you get a lot of artifacts when encoding interlaced material . Please refer to the documentation of the filter for further information.

/Mats

[kamaleon]

How good is AutoGK's deinterlacing compared to TMPGEnc's?

Probably better. While you're doing all of this writing and asking of questions, you could have been running some practice encodes. Reencode some videos. Try out AutoGK and/or TMPGEnc for the job and decide for yourself. One thing, though; AutoGK will make all the hard decisions for you, and make them right. You on the other hand, if using TMPGEnc together with your apparent lack of knowledge, are more than likely to make some wrong ones.

Wow that's a bit harsh innit? I appreciate you trying to help, but actually letting AutoGK do everything might not really teach me much...

I'm giving it a try anyway

[kamaleon]

Pffuah..

AutoGKresized the image to 720x544, cropped some black bars but:

Doesn't allow to specify XviD's bitrate (correct me if I'm wrong), created a 25fps video which is not fluid at all (I'm encoding football matches) and its deinterlacing looks disgusting to my eyes, full of artifacts:



As you can see there's nothing of the sort in the video encoded by TMPGEnc.

I maybe a noob, but I'm looking for good quality, so any help is appreciated.

[guns1inger]

You specify either a desired filesize (which is the same as specifying a bitrate anyway - filesize = running time x bitrate) or or you can do a quality based encode instead.

[Midzuki]

I maybe a noob, but I'm looking for good quality, so any help is appreciated.

Why not use VirtualDub and start learning some Avisynth
We all were noobs "sometime before 1975"

[kamaleon]

You specify either a desired filesize (which is the same as specifying a bitrate anyway - filesize = running time x bitrate) or or you can do a quality based encode instead.

Yes, I understand that. I suppose it's just an indirect way (and a bit more awkward) way of specifying the bitrate by having to calculate the final file size. I'd rather not have to do that, but ultimately I don't think that's the main problem I see in AutoGK

Why not use VirtualDub and start learning some Avisynth
We all were noobs "sometime before 1975"

I knew someone was gonna say something of that sort
Seriously, it's too much trouble for me to learn all of that just to encode some videos to gain some space. I'm sure there's something between the total noob AutoGK do-everything-for-a and the incredibly complex avisynth hey? 8)

I mean don't get me wrong, I'm not lazy and I'm not waiting to be spoonfed, but very honestly, I really can't get around to the quintessence of video encoding using avisynth etc...

Now, to get back to the subject,

In what regards 25fps vs 50 fps,

My eyes tell me that the image is not fluid @ 25 fps, is there any downside to encoding @ 50 fps?

[kamaleon]

KernelDeInt filter is a filter performing deinterlacing in AutoGK. Its threshold parameter allows you to control sensitivity of the filter in detection of motion areas (as static areas are ignored by the filter). This is very advanced settings and it should be used only when you get a lot of artifacts when encoding interlaced material . Please refer to the documentation of the filter for further information.

Yesterday I was giving tdeint a try, along mvob (couldn't get mcbob to work, keeps giving an error), but I can't really come to grips with scripts and avisynth. I'm really struggling with this. I don't think I can really use these filters, and specify thresholds if I'm not using a WYSIWYG GUI.

The thing is, I've been using nvidia graphic cards with the purevideo processor, and I've never been able to deinterlace a video with the same quality as I can see purevideo perform during playback.

http://www.sendspace.com/file/h2viid

Here I've uploaded a couple of screenshots of different deinterlacing I've seen mpeg2 codecs perform, If someone could tell me if it would be possible to encode to XviD achieving a result similar to what Dscaler seems to perform using a purevideo nvidia card, you would make me someone *really* happy.

[guns1inger]

My eyes tell me that the image is not fluid @ 25 fps, is there any downside to encoding @ 50 fps?

Yes, but if the source is PAL, you gain nothing and may even lose. Options are

1. Split the fields. You get 50fps, but only half the vertical resolution

2. Duplicate every frame. Again, you get 50 fps, but the motion won't be any smoother

3. Use complicated motion estimation techniques to create a new frame between each current pair. To do this would take complicated avisynth (and I mean complicated, not the basic resize stuff you are already scared of) and processing will be slow. And again, little guarantee that it will in fact make it any smoother.

Your source is 25 fps. Does it look smooth ?

[mats.hogberg]

You could also try avi.net - takes some other routes than AutoGK on the way from mpg to AVI.
Or look here https://www.videohelp.com/tools/sections/dvd-to-avi-divx-xvid for more tools. But in the end - Learning AviSynth and/or VirtualDub may be the only way to arrive at better quality.

/Mats

[guns1inger]

Forgot option 4

Speed the whole thing up to 50 fps and finish the game in half the time

[kamaleon]

My eyes tell me that the image is not fluid @ 25 fps, is there any downside to encoding @ 50 fps?

Yes, but if the source is PAL, you gain nothing and may even lose. Options are

1. Split the fields. You get 50fps, but only half the vertical resolution

2. Duplicate every frame. Again, you get 50 fps, but the motion won't be any smoother

Do you know which method TMPGEnc uses? When I set it to 50fps everything seems fine to me. Am I loosing something in that case?

3. Use complicated motion estimation techniques to create a new frame between each current pair. To do this would take complicated avisynth (and I mean complicated, not the basic resize stuff you are already scared of) and processing will be slow. And again, little guarantee that it will in fact make it any smoother.

You are right - I am scared - I'd rather not have to think about that solution

Your source is 25 fps. Does it look smooth ?

I don't know the answer to that... isn't it actually that 25 fps interlaced = 50fps progressive when deinterlaced? I mean, using the right mpeg2 codec yes it does look smooth, but is it really 25fps?

Forgot option 4
Speed the whole thing up to 50 fps and finish the game in half the time

[guns1inger]

Interlaced means you get 50 fields per second, not frames. Depending on how you deinterlace you may get 25 frames per second, or you may get 50. If you source is PAL then you get 25 interlaced frames per second when you play it back.

[kamaleon]

Interlaced means you get 50 fields per second, not frames. Depending on how you deinterlace you may get 25 frames per second, or you may get 50. If you source is PAL then you get 25 interlaced frames per second when you play it back.

Ok, so what's responsible for the lack of smoothness when I watch an xvid in 25p vs mpeg2 25i ?

I mean, mpeg2 using dscaler, nvidia, cyberlink, etc, all give smooth playback, but when I encode to xvid @ 25 fps ( I suppose that makes it 25p right) it's not watchable?
If I choose 50 fps in TMPGEnc it looks wonderful though. What's the catch here?

[manono]

Granted that AutoGK pic you posted doesn't look so good, although I daresay that at full motion you wouldn't notice the artifacts. However, if you encoded the TMPGEnc pic at 50fps, then you made an apples to oranges comparison. It's unfair to both to try and compare them. In addition, I'm pretty sure that you'll lost half the vertical resolution using TMPGEnc's bobber when you make the videos 50fps. Although with sports, where the motion is pretty constant for both camera and players, just about any deinterlacer will lose resolution. If, for example, you went and found a thin diagonal line, like a white field boundary line as seen from a distance, I'm pretty sure the TMPGEnc frame would look pretty bad. I'm not real sure how the AutoGK one would look, though, maybe just as bad. Sports are among the most difficult kinds of video to deinterlace. Here's a deinterlacer comparison. Only the very best deinterlacers (and the very slowest) can even begin do sports videos justice:

http://heptium.sh.cvut.cz/~integra/deint/

Ok, so what's responsible for the lack of smoothness when I watch an xvid in 25p vs mpeg2 25i ?

Having half the original framerate. I don't know enough about the dscaler/NVidea/Cyberlink deinterlacers to make comparisons, but if you want smooth, then try a simple FieldDeinterlace() in your script (It's part of Decomb). It'll play smooth - and blurry.

[guns1inger]

Post a couple of short clips - one from each encoder - so what can see what you are creating.

[kamaleon]

I daresay that at full motion you wouldn't notice the artifacts.

I wish I couldn't. They're all over the place, trust me. I'm very sensitive to artifacts. I spend a lot of time watching sat tv on my computer and I've developed a pretty nasty habit of seeing jaggies, ghosts, blends, and artifacts. But in this case, they're really blatant, believe me. Funnilly enough, the last time I've seen some mpeg2 on an ATI card those where exactly the same artifacts. I wonder if they use the same techniques that kerneldeint does.

I'm pretty sure that you'll lost half the vertical resolution using TMPGEnc's bobber when you make the videos 50fps

Care to elaborate? I lost resolution? What does that mean? Image looks good to me...

Here's a deinterlacer comparison.

My god, securedeint, tdeint and mvbob look pretty good - apart from those blatant white lines on the upper half. Yeah, sports are tough hey - especially those with white lines on the field

Ok, so what's responsible for the lack of smoothness when I watch an xvid in 25p vs mpeg2 25i ?

Having half the original framerate.

Yeah this frame / field thing does my head in, really. So is there any deinterlacing that keeps the original smoothness without loosing vertical resolution (still didn't find out what that means, or if that's bad)?

[kamaleon]

Post a couple of short clips - one from each encoder - so what can see what you are creating.

No problem!
http://www.sendspace.com/file/94o002

Nevermind about the bitrates, the cropping etc, I wasn't paying attention to that. Only to the deinterlacing / fps issue.

FYI, the original mpeg2 clip is in the archive I've posted earlier on, along with screenshots of the deinterlacing done by different codecs.
You should note that TMPGEnc seems much better than AutoGK but still doesn't compare to any codec that uses nvidia's purevido hardware as I've outlined in red