Reading MPEG2 (muxed video+audio) into Avisynth

[voodle]

...eg, any standard MPEG2 file from, say, a digital video tuner.

Lots of (different) advice about the best way to do this. DGIndex seems to be a favorite, but (question coming up)...

DGIndex is slow, and requires a separate process that you need to run before loading into Avisynth. It creates separate video and audio files which you have then to mux back together with AudioDub().

(And here's the actual question at last)

Why do folks find this preferable to pulling the MPEG2 into VirtualDubMod, setting up an avi frameserver (a cinch in VDubMod), and then reading the (fake) avi directly into AviSynth with AviSource()?

OK, you're still running a separate process, but you don't have to mess around demuxing, creating extra files and then muxing them back again.

I ask this question with the deepest respect, because I know smart people (and smart apps, like AutoGK) are using DGIndex, so there must be a good reason. Is there some problem with audio syncing in VDubmod I don't know about?

--
voodle

[guns1inger]

If you install FFDShow and set it to decode mpeg-2 files you can then open them in avisynth use DirectShowSource()

Generally I would be doing this to filter, and avisynth has many advantages, including colourspace handling, and more importantly, speed. Many filters in avisynth are substantially faster than vdub equivalents, so even with DGindex the avisynth process is faster.

[voodle]

If you install FFDShow and set it to decode mpeg-2 files you can then open them in avisynth use DirectShowSource()

I believe DirectShowSource will open MPEG2 files with or without FFDShow, won't it? It just uses the built-in M$ reader. And is to be avoided, in general, because its loose behaviour tends to produce out-of-syncness.

I take your point about VDub filters, but of course in this instance I'm not suggesting using any VDub filters. Just reading the MPEG2 file into VDubMod, setting it up as a (fake avi) frameserver and banging the output straight into Avisynth using AviSource.

So what's the advantage of DGIndex?

--
voodle

[manono]

...so there must be a good reason.

Sure. How about what you're suggesting is old, obsolete, slow, and produces inferior results? To avoid that you'll begrudge the 2 minutes or so to create the D2V Project File?

http://avisynth.org/index.php?page=Section+1%3A+About+AviSynth#q1.12

I don't use AudioDub, by the way, but keep the audio separate and mux it back in after encoding. Maybe you have a reason for keeping the audio with the video.

[voodle]

what you're suggesting is old, obsolete, slow, and produces inferior results

Sorry not to have been clear, but I'm trying to find out why and in what way frameserving through VDubMod produces inferior results. Is it a colourspace issue, or what? Is a frameserve from VDubMod really slower if you're not using VDub filters?

Maybe you have a reason for keeping the audio with the video.

I need to AudioDub the two streams together because I use AviSynth (+AvsP) to do my video editing (basically Trimming and Dissolving). Mute video tracks can leave you with misedits, even if you can manage to match up the audio afterwards accurately.

--
voodle

[manono]

I thought the link would have explained why it's both slow and lousy to do it your way. It's slow because of the reasons guns1inger explained earlier. AviSynth filters are faster usually. Yeah, I know you said you're not doing any VDub filtering. Plus, by keeping it in the original colorspace, it's way faster. Have you ever compared the time difference? They're not kidding when they say it's up to 35% faster to keep it YV12 all the way, if possible.

For quality reasons, you don't want to convert colorspaces any more than absolutely necessary.

[voodle]

OK, this is very helpful. Thanks to manono and guns1nger.

I see that when VDubMod is used as a frameserver it delivers RGB24, which I understand is a conversion from MPEG2's native YUV12 colorspace. I can see why this is undesirable, thanks, particularly in the light of the explanation here: <http://www.animemusicvideos.org/guides/avtechbeta/colorspace.html>

I've typically been using VDubMod to feed the stuff into XviD -- I understand that this will still preserve YUV12, as long as I stick to Fast Recompress for my second pass. Does that seem right?

I'd love to be able to get AVS2AVI working properly from inside AvsP, but having set up AvsP to find my copy of avs2avi.exe, I still can't get it to do anything more than just say Done, without actually doing anything. Any hints on that would be welcome.

--
voodle

[voodle]

OK, this is very helpful. Thanks to manono and guns1nger.

I see that when VDubMod is used as a frameserver it delivers RGB24, which I understand is a conversion from MPEG2's native YUV12 colorspace. I can see why this is undesirable, thanks, particularly in the light of the explanation here: <http://www.animemusicvideos.org/guides/avtechbeta/colorspace.html>.

I should add that there's one other reason, actually the main reason, I've been using VDubMod. I don't know any better way of getting rid of Channel Logos (eg More4) than Karel Suhajda's VDubMod DeLogo filter. It can do a complete vanishing act. So I've been using VDubMod+DeLogo as a frameserver into AviSynth, doing trimming and dissolving in there (mostly to get rid of the ads) and then frameserving the AviSynth output into a second instantiation of VDubMod to drive XViD. If somebody knows a way of accomplishing the same thing while staying in YUV12 I'd love to hear about it.

It would be nice, too, to be able to get AVS2AVI working properly from inside AvsP, but having set up AvsP to find my copy of avs2avi.exe, I still can't get it to do anything more than just say Done, without actually doing anything. Any hints on that would be welcome.

--
voodle

Sorry, guys, I've absolutely no idea how this self-quote got in here. And, sorry, I can't see any way to delete it.

Well... the topic is called "Newbie Conversion". Guess I've failed to be converted.

--
voodle

[manono]

Nice article you found. I know 3 of the 4 authors, and they're big guns in the AviSynth world. I didn't know syskin was an anime fan, though, although the other 2 I know are. The article is well written and accurate.

I've typically been using VDubMod to feed the stuff into XviD -- I understand that this will still preserve YUV12, as long as I stick to Fast Recompress for my second pass. Does that seem right?

Yes, you'll keep it in YV12 by using Fast Recompress for the 2 passes.

If somebody knows a way of accomplishing the same thing while staying in YUV12 I'd love to hear about it.

Not possible. To use any VDub filters in an AviSynth script, you have to convert to RGB24. More information here:

http://avisynth.org/index.php?page=Section+4%3A+Importing+filters+from+VirtualDub

There is a speed advantage to doing all the filtering you can in AviSynth, before then using the VDubFilter. You're converting to RGB24 first thing, and then doing your filtering.

There is one AviSynth native delogo filter, XLogo. I have no idea if it's any good. You can read about it here:

http://www.doom9.org/capture/delogo.html

and find it here:

http://avisynth.org/warpenterprises/

Any hints on that would be welcome.

I don't know anything about AVS2AVI or even AvsP. Maybe someone else can help.

[voodle]

Thanks for all that, manono. I did try XLogo, but IMHO it's not nearly as good as DeLogo, which can be near-perfect if used right on a semi-transparent logo.

I don't know anything about AVS2AVI or even AvsP. Maybe someone else can help.

It's worth investigating AvsP, manono. IMHO it's a terrific AviSynth editor.

--
voodle

[Abond]

It's worth investigating AvsP, manono. IMHO it's a terrific AviSynth editor.

Well, it is not completely correct, the editor IS avisynth, AvsP is more like a GUI with "live" preview. But I think it can be very useful. Unforttunately I have no time (and patience maybe ) to play with it

[voodle]

Well, it is not completely correct, the editor IS avisynth

Gosh, one really does have to watch ones ps and qs here.

AviSynth is the video editor. AvsP is a text editor for Avisynth scripts, with many bells and whistles.

(Er, when I say "bells and whistles", I don't mean it literally whistles.)

--
voodle

[Abond]

Yeah, but it definitely "bells"

[voodle]

And I see I did call it an "avisynth editor", not a video editor.

Just to put this topic to bed (and, thanks, I've learnt a lot), let me ask one more question. There's no actual loss of colour gamut, is there, in the conversion from YUV12 to RGB24 that VDubMod does?

If I'm right about that, the downside is just:

1. processing will take more time (but CPUs are getting cheaper and faster by the day), and

2. Compromises and bugs in the implementation of colour conversion may introduce colour errors (but these may well be minimal).

What does the team think?

--
voodle

[manono]

I think your 2 points are a fair assessment of what happens. As your article states, the color errors of your point 2 might take the form of a slight blurriness and a slight fading. The more colorspace conversions that take place, the more likely these errors are to be noticed.

Sometimes, as when you want to use a particular VDub filter, it can't really be helped. Other times, as when you use your method for editing your captures, it's just more convenient for you to do it your way, even though it's possible and not all that difficult to do it in such a way as to avoid the colorspace conversions. And sometimes, out of ignorance, or an unfamiliarity with AviSynth, people do it the wrong way because they don't know any better.