Recording Video: What's Best to do?
 

I captured a 3hr sportin event and I was wonderin which is best to do??
1. use the 22gb m-jpeg file and use that straigh through tmpgenc
or
2. convert it to divx andnthen encode to kvcd

which one would anyone recommend?? and what is the best script to use??

Re: Whats best to do???
 

A,
there's really no need to do divx (unless you want to, say for archival purposes), you can go directly to kvcd. If you haven't already, u may wish
to download the kvcd samples, burn them to cdrw and test them out on your dvdplayer to make sure it can play or which kvcd resolution rather, your dvdplayer can play.
For scripts, it'd depend on the quality of your capture, noisy, vhs, satellite, etc...but we can work on scripts.

ztr

well its analogue capture from satalite tv, pretty clean capture through vfw drivers. I have done many dvd to kvcds but this would be my first attempt at avi to kvcd so any help will be great lol.

andybno1,

You can do that for educational purposes, otherwise I advise against it. To get the best results, MJPEG is the way to go.

[]'s
Vmesquita

ok I'l go with mjpeg then, so seein as I don't have avisynth .20 anymore is there a 2.5 script for avi to kvcd?

There is an optimal script for VHS, and I can give the one I use depending of your source... What is your source? VHS, Satellite, cable...? :D

[]'s
Vmesquita

I tried the optimal script but got a load of not 2.5 compatible filters so gave up lol, the source is a clean stalite capture through vfw drivers.

so any 2.5 scripts??

Here I posted an AVS 2.5 capture Script, I think your capture is interlaced so it also contains a good adaptive deinterlacer, ... but also see the commends further down of Jorel and DialHot for optimisations ;-)

http://www.kvcd.net/forum/viewtopic.php?t=5067&start=16

It's your luck day!

I have a very good script using undot and convolution3D adapted for interlaced stuff! I use with all my satellite captures and they look great. I'll post later (I don't have access right now) for you to try. :D

[]'s
Vmesquita

cheers for the help will try that scriptout and await vmesquita's script to try out aswell.

Do send.

Sorry for the delay... Here it goes:

a=AviSource("capture.avi")
a.undot().SeparateFields()
even=SelectEven(a).Convolution3d(0, 10, 17, 10, 13, 2.8, 0)
odd=SelectOdd(a).Convolution3d(0, 10, 17, 10, 13, 2.8, 0)
a=Interleave(even,odd).weave()

a=Crop(a,24,16,680,448).AddBorders(24,16,0,16).lan czosresize(352,480)
return(a)

[]'s
Vmesquita

Which template/resolution are you using for encoding?

When doing interlaced stuff for DVD, I always use 352x480.

[]'s
Vmesquita

firstly heres the script I used as indicated in the link posted by incredible:

and well I got one interesting colour sceme.

here is the sample. Anyone know what is wrong here?

(first you have to tell the people to copy the URL in the browser adressbar directly to get into the link, it seems that Yahoo prohibits links to Files)

Well your encoding example seems to be strange!
Maybe its a Problem when using this script handling NTSC Streams?

As I told ....

.. this line has to be used with caution, cause I don't know if these Values are the same for NTSC.

But, how does the Pictore look like when opening the .avs in Windows Media Player?
ok?

If there are still Problems ... try to switch off these lines as quoted right here above by adding a "#" in front of the lines and do a test, also by watching the .avs in WindowsMediaPlayer.
If the problem is gone, then you have to correct the Number Values of the Vdub PlugIn by loading this movie into Virtualdub, adding the SmartDeinterlace Filter an do your special settings in the Window.
Let it run in Vdub to test your deinterlacing. If its ok, safe the processing Settings in Vdub and open the .vcf file using Windows Editor.
Copy exactly the new values strored in the line where the SmartDeinterlace command is shown and post them into the capture .avs script.

Another problem could be the ConvertToYV12() command.
Whats the colorspace of your Avi?
Try to switch off the ConvertToYV12() line and do a test. Gone?
Or try to change it to ConvertToYUY2().

Well lets see if it helps.

Let me know .... ;-)

incredible,

Avisynth has such great deinterlacers, why use an old VirtualDub plugin to do the job :?:
:idea: Use one of the great VirtualDub Deinterlacers, like Tomsmocomp (my favorite), fielddeinterlacer, GreedyHMA... :wink:

[]'s
Vmesquita

Hi all..

just jumping in here.. what the heck.. :hihi:

In my script-of-things for my encode projects, I have yet to find an AVIsynth
(AVS) script that handles Interlace well enough for me to accept. I mean,
I've about tried them all, and nothing seems to come close to the level of
quality output that I can, have and do acheave when using vdub filters.. or
should I say, de-Interlace techniques.

But, remember this one important fact which everyone here, there and pretty
much everywhere else is missing..
That Interlace sources are NOT one in the same. They vary from one to
another, for reasons (at the moment) beyond me. What I'm saying here, is
that there seems to be more than one variation of Interlace Types.

My conclusions are based on the many different de-interlacing techniques
I had to employ to many sources, and even though they were all Interlace
sources, the same de-interlace algorythm or technique or filter did not always
produce the same output quality. I find myself having to resort to various
de-interlace techniques, hence the need to index them. And, another thing
to mention here on Interlace..

Well, the above was my two cents worth.
-vhelp

Hi Vmesquita! ;-)

I tried a lot of avisynths deinterlacers but the best for me is Deinterlace_MAP but too slow even on my overclocked athlon 1800@2400Mhz real. So SmartDeinterlacer gives the best compromoise of quality and speed to me. I tried Fielddeinterlace from avisnth but it seems that its a Non-adaptive deinterlacer, anyway the quali of smartdeinterlacer was better cause of its adaptive deinterlacing routines.
But ... Ill try it again and also the Greedy one, TomsMoComp I know also.
;-) Thanks for the tip. ... Ill run next weekend the big-deinterlacer hall of fame test ;-)))


@vhelp

Youre exactly right! Thats wy I first modify the settings during a test in Vdub, like I explained in the reply above. There are many techniques like even odd field techniques and so on.
Well the best is to find the best deinterlacer you stay with and integrate it into the capture script above ;-)

well for starters I didn't know geocities had a link protection on as it was fine when I left clicked it.

I'll have a go at those lines, bt I get same results in media player. and this aint a ntsc stream its pal capture from satalite. Nothin wrong with my encodin samples, its only that small cause I have dialup and decided to only do a few seconds encodin. The colourspace is 16 rgb

but anyway I'll get onto this one way or another lol.

Satelite???

How do you capture? Directly from the receiver?
Do you have the possibility to capture digital?
If yes, this stream could be progressive if its "captured" (better said "streamed") digitally.

Well give me infos ;-)

But anyway if you receive a PAL program via satelite, your receiver will convert it to NTSC so your TV will accept it. So you still receive a NTSC interlaced stream from your receiver if captured using an analog capture-card.

its a satalite feed captured via vdub in vfw drivers. captured at 640 480 (don't ask why) usin mjpeg codec with forecd quality of 80and capture quality set to 100 colourspace is rgb 16 bit

Whats the name of your card?
Which mjpeg codec do you use?
and capturing using rbg 16bit???? Wow, isnt it possible to capture at least at rgb 24 or even YUV??

I think there is a problem with the colorspace you use, and it seems that you really capture at NTSC interlaced, thats why you have to check the script as I mentiones above (deactivating the deinterlacing lines just for testing).

Well the standard capture size for PAL is 704x576 and I think for NTSC its 640x480, .. thats why.

Try to use the PicVideo Codec for capturing, it gives you a good quality by using less HD space for capturing. And you can capture using YUV!

could ya give a link to this codec pls.

http://www.pegasusimaging.com/picvideomjpeg.htm

Its a trial version and it puts a note on the black bars in the screen, but you can overlap it when using the Avisynth command

Letterbox(xx, xx, 16, 16)

Or you use the HuffYUV codec:
http://www.divx-digest.com/software/huffyuv.html

It needs more space on HD but also as named a YUV Codec ... and its free!

@andybno1,

According to someone at dvdrhelp.com, you can also install the Trial of ShowShifter (a TV recording program) and deinstall, and it will leave behind a freeware Picvideo MJPEG codec. Never tested but lots of people said it works.
And for really lossless codecs, besides huffyuv, there are some lossless YV12 codecs in doom9 forum that take about 70% less space than HuffYuv. If I remember correctly, the name of the codec is VBLE. I have used it with great results, but generally I use PicVideo MJPEG because I only got a 40 Gb HD! :(
Capture using at least YUYV, any RGB is a waste of space and a consequent loss, since your source signal is YUYV, and you're converting to RGB.

@incredible and vhelp
I like avisynth deinterlacers and they don't perform slow in my atlhon XP 2000+, but It's also a matter of taste... :D For trully interlaced stuff I prefer Tomsmocomp or FieldDeinterlace(blend=false). GreedyHMA is said to be good with hybrid material. By the way, decomb 5.00 (which fielddeinterlace is part of) has been released this week, I haven't tried it yet but you can find at doom9 forum...
Actually, I don't like to deinterlace, because the picture looks wired for my eyes when there's movement. Even when back in my CVD days, I never deinterlaced stuff. I believe there's no way to build a perfect deinterlacer since you need a temporal information you don't have... :D But that's of course my personal oppinion.

[]'s
VMesquita

Ill try it again to see ;-)

I totally agree with you!
But with noisy captures you already get less CQ (even when handling them via Avisynth) and so I decided for myself to work with adaptive deinterlacing methods like the SmartDeinterlacer to be able to encode at mpeg1 which only supports non-interaced encoding. Ther're also other adaptive deinterlacers and I'm still running through the web to find the best for me which gives me the Quality I need. The advantage of deinterlacing is also to encode your mpeg afterwards in mpeg1 which doesn't support interlaced encoding.
So I'll try FieldDeinterlace(blend=flase) to compare it with the smartDeinterlacer and others ... ;-)
And thats the point of captures: If you work with DVD mpeg sources you most will work with a good quality sources, but in case of captures there is a mega-big bandwith from noisy bad up to good little-grainy.
In my case of handling captures with avisynth, every avisynth script for each capture is special "modifyed" cause of the difference of each capture.
(more less denoising, more less sharpening and so on...).
There will not be the "right" capture avs script to use for general, its just a basic script as a base for further optimizations.

;-)

So lets see Varmesquita if we get a good combination of our both experiences. Would be great ... cause it seems that you do the "other" way: NTSC ... right?

Tantos saludos
Inc.

incredible,

There are pretty good (and pretty slow) denoising plugins for avisynth. For real hard analog noise, there's nothing better than GoldDust (from the dust package), GrapeSmoother and CNR. But of course you have to regulate youre filtering according to the level of noise. To use GoldDust with analog stuff I have a special script, I can give you if you want.
For not so hard analog noise, PeachSmoother, CNR and GrapeSmoort are my choice.
For my Satellite captures, I use almost always the Convolution3D script I posted. The only modification I sometimes do is remove Undot when the signal is very clean.
As vhelp pointed, some stuff are really worse to deinterlace. The worst material is those shaky and panning home movies. They will look awful deinterlaced, no matter what you do. But I deinterlace a ABBA DVD and it looked OK most of the time.
An extra tip. If you want, try Procoder (an MPEG encoder) using MPEG2 CVD at 352x480. Encode interlaced, 1-pass VBR in Mastering Quality mode. Beware that's very slow, but it'll produce the best output from interlaced noisy stuff you'll ever see.
And yes, I am in NTSC land! :D

[]'s
VMesquita

well I have found out its not any of the filters, I done a script with just the avisource line in and I still got the colour problem so god knows whats wrong.

Could you post some screenshots of the color issue? I couldn't get the mpg file in that link earlier in this thread.

right click the link and save target as just tried it then and it works for me.

Looks to me you'll need SwapUV() added to the script. Put it right after the AVISource line.

cheers for the SwapUV() line worked a treat. :ole: hurray for Boulder :ole:

Hey! I'd like to have a look at your noise reduction scripts. I'm also capturing NTSC VHS as 352x480 Huffyuv. If you would, post 'em up or PM me. :D

vmesquita,
I too would be very interested in seeing your noise reduction scripts using GoldDust. NTSC VHS @ 352x480 Huffyuv. Also, is there a filter that will help control the 'blooming' of red colors in a VHS capture?

Ok! I'll post it later. :D
I think there isn't such a thing... :( The only way to cure this would be using a TBC.

[]'s
VMesquita[/quote]

You might be able to do something about the colors by using RGBAdjust (internal Avisynth 2.5 filter).

From the documentation:

RGBAdjust(clip, float red, float green, float blue, float alpha)

This filter multiplies each color channel with the given value, clipping the result at 255.

You could try reducing the red channel and see if it helps.

Here it goes:
I resize before filtering because GoldDust is very slow. You can Tweak GrapeSmoother according to the source. Please remember that this is only for very Bad VHS.

[]'s
VMesquita