AVI Capture from Satellite to KVCD?

[e1miran]

I've been trying to get the "best" looking KVCD of some programming that I recorded on my DishNetwork PVR (like TiVO). I captured it to my PC at 352x480, YUY2, 29.97fps. This is full-screen, interlaced material, of course, and it's about 75 min long.

I also would like 75 min to be my max length for most of the programming that I capture (2 discs will hold 2.5 hrs - enough for movies from TV)

1- What are your suggestions for encoding with TMGEnc? Should I use the "KVCDx3-MPEG-1-NTSCFilm" template - which is what I've been using? This template is setup for 528x480 output and I've experimented at re-sizing. Or should I be using the "KVCD-CQ-352x480-_NTSCFilm" template?

2- I read in this forum that the templates and scripts are primarily for widescreen not fullscreen. So, what changes to the settings or avisynth scripts should be made for fullscreen source?

3- What is the purpose of using gripfit(gripcrop)? It's not needed if the source and output are both 352x480, right?

4- What suggestions do you have for the avisynth script? I need to deinterlace and convert fps in the script, right?

I know that I'm asking a lot of questions. And yes, I've been reading up on this for a few weeks now. I've done my own experiments, but am still a little ignorant to what some settings are for and the best filters or order of filters to use in avisynth.

Thanks in Advance.

[Dialhot]

1/ you can use any template as they basically differ only in the resizing and in the min/max video bitrate.
But you have to use a script that will deinterlace correctly the source. In the optimal script section of the forum you will find this :
http://www.kvcd.net/forum/viewtopic.php?t=8678

The script also take in charge the resolution, that is why any template can be use : it is not the template that control the resolution, but the script.

But in your case, see point 3 just below.

2/ The template are for both but the results we announce (120 min of video on a CD) are irrelevent if you are in full screen. The KVCDx3 for instance allow to put 100 min in windescreen but ony 75 in fullscreen.

3/ you are right : in your case you can remove all part of the script that concern the resizing. In this case, do control after you load the source in tmpgenc taht the resolution is correct (always load the template first, then the avs script. This way the res will be okay).

4/ already answered

[e1miran]

WOW! What a fast response. Thanks. Any other suggestions as to get the most out of KVCD for my situation is appreciated. Unless of course Dialhot said it all.

Thanks!!

[Dialhot]

As your video are 75 min long, if you can capture to a greater res it will be better. Because in 352*240 I can put easily 2h30 of video on a disc. So for 75 min, I think that you will have empty space on the CD

[e1miran]

Quote:

Originally Posted by Dialhot

As your video are 75 min long, if you can capture to a greater res it will be better. Because in 352*240 I can put easily 2h30 of video on a disc. So for 75 min, I think that you will have empty space on the CD

I'm capturing at 352x480. Is that twice the resolution vs. 352x240? In my case, capturing at more than 352x480 leads to unacceptable amounts of drive space being used - Huffuv codec.

[Dialhot]

352*480 will give sharper (and nicer looking) image. So use it if you can. Of course your KVCD will also be 352*480.

[e1miran]

Phil,

Since my source is AVI that I captured, should I use your AVI Script as opposed to the KWAG's "Optimal Script"?

If it's better to use a combination of the two, what would it look like?

Thanks in Advance.

Eli

[Dialhot]

Quote:

Originally Posted by e1miran

Since my source is AVI that I captured, should I use your AVI Script as opposed to the KWAG's "Optimal Script"?

No.
We often write "avi->kvcd" script when we talk about the scripts I did but they are actually "MPEG4 -> KVCD" scripts.
Here you are using Huffyuv as avi codec. This is a lossless codec and your image is as pure as it would be if you didn't compress it at all. They are far closer to a DVD than a MPEG4 (Divx, Xvid...).

Quote:

If it's better to use a combination of the two, what would it look like?

Nothing really interresting. Each script is optimized for what it is supposed to do. Ketchup is good, chocolate is good; But I'll never try Chocolate + Ketchup together

[incredible]

First choosing at least 352x480 is a good choice as you preserve the fielbased architecture by doing this (also full width would give more sharpeness but ...ok).

Look at the capture content ... is it a movie, means originally shot on FILM material? Yes? then:

Avisource("D:\The Path\To your\capture.avi")
Telecide()
Decimate()
Bicubicresize(...........)
Deen("a2d",1,5,8).Temporalsoften(2,5,8,15,2)# just a try but you can lower the values
Addborders(......)

If its not a movie but a documentation, reportage, etc., ... means shot originally on VIDEO material then:

AviSource(""D:\The Path\To your\capture.avi") 
bob() # or "SmoothDeinterlace(doublerate=true)", also Sharfis Brains new "BobMatcher()" is fantastic as all do output a 60fps! deinterlaced stream
convert60ito24p(2,0) # The function below called 
BicubicResize(...........) 
Deen("a2d",1,5,8).Temporalsoften(2,5,8,15,2)# just a try but you can lower the values
Addborders(......)
# 
############### the "convert60ito24p" function by sharfis_Brain  ############# 
# 
function convert60ito24p (clip video, int mode, int offset) 
{ 
work = assumefieldbased(video) 
out = (mode==2) ? interleave( 
\selectevery( 
\layer(trim(work, 1, 0), 
\layer(work, trim(work, 2, 0), "fast"), 
\"fast"), 5, 0 + offset), 
\selectevery( 
\layer(work, trim(work, 1, 0), "fast"), 5, 3 + offset)) : 
\* * (mode==1) ? interleave( 
\selectevery(trim(work, 1, 0), 5, 0 + offset), 
\selectevery(layer(work, trim(work, 1, 0), "fast"), 5, 3 + offset)) : 
\* * (mode==0) ? selectevery(work, 5, 1 + offset, 4 + offset) : work 
assumeframebased(out) 
} 
#################################

In the case above Bob() is used to deinterlace to 60fps, but Bob() ist "stupid", means he does something "like" seperating the fields and afterwards resizing them to full height!
So u should use a smart bob like Smoothdeinterlace(doublerate=true) or even the best now on the script market (IMHO) Sharfis_Brain's new BobMatcher() (look out at forum.doom9.org for the code).

You also can try to change from HuffYUV to a the better compressing "PicVideo mjpeg codec" as its also very fast and does output if using "quality=19" a very good quality in also YUY2!

[e1miran]

Incredible,

Thanks. Source is a movie - Disney's "Dinosaur".

A couple of questions--of course. Why is there no deinterlace filter for the "Film" source in your first script? Is this what "telecide()" does? From what I understand even if it was originally film, it is now interlaced for TV. I do get combing if the AVI is viewed alone.

Is there any advantage to running the fps conversion function with the first script? If not, why?

As noted above, my output res will be the same as my input res. So, I do not need to resize nor add borders, right?

Finally, these scripts you provided are optimized for capped material, so I shouldn't mess with KWAG's "Optimal script" nor include the MA script settings when it comes to this kind of source -- is that your recommendation?

And I do have the PicVideo MJPEG, but in my opinion the HuffyUV capture looked a lot better even with MJPEG set at 20.

Thanks!!

[incredible]

Quote:

Originally Posted by e1miran

Incredible,

Thanks. Source is a movie - Disney's "Dinosaur".

A couple of questions--of course. Why is there no deinterlace filter for the "Film" source in your first script? Is this what "telecide()" does?

Your "Movie" isnt really interlaced as its not really recorded at 240px heigth and 60fps! These Fields will be weaved and that gives 29,76 fieldbased fps!

But your movie was shot on film (let me say more: it cant be "originally" interlaced cause it bases on full height painted "FRAMES"!

So what do they do to get that stream ready for NTSC broadcasting?
Very easy, they do a Pulldown to get 29.97 fps out of 23.976 (film!) means frames will be added to result in that higher fps. Afterwards a phaseshift or dynamical phaseshift will be performed so the "fileds" will be moved by for exmple one field and if these fileds get waved again, it "appears" interlaced.

If you add in your progressive source based avs script a:

separatefields()
Trim(1,0) # shift 1 field
Weave()

Thats would end up in that "like" interlacing look.

So thats what telecide() does: It restores the orig Frames! out of the fieldshifted weaved fields. = back to the original progressive state.

Decimate delets out the doubled frames of the telecined broadcasted stream.

Quote:

From what I understand even if it was originally film, it is now interlaced for TV. I do get combing if the AVI is viewed alone.

Is there any advantage to running the fps conversion function with the first script? If not, why?

NOPE!!! You already do restore the stream to 23.976 by using telecide(), decimate() means IVTC = inverse telecine! CAUSE IF THAT MOVIE BASES ON FILM MATERIAL then its NOT really interlaced! And we only do perform 60ito24p on REAL interlaced sources which are shot on VIDEO material

Quote:

As noted above, my output res will be the same as my input res. So, I do not need to resize nor add borders, right?

You should get rid of the broadcasted borders and tweak the colors/ luma as mentioned in my capture guide in the capturing part of this forum.
There you also see how to crop the borders and how to resize correctly.

Quote:

Finally, these scripts you provided are optimized for capped material, so I shouldn't mess with KWAG's "Optimal script" nor include the MA script settings when it comes to this kind of source -- is that your recommendation?

Caps do often got a lot of noise and MA only performs a temporal denoiser on very slow scenes! Means in case of caps if your image content begins to "move" you will end up with more "blurred" noise but noise is still visible.
MA is very well optimal for good DVD Sources.

You can Tweak it by insertin a Deen("a2d",1,5,8 ) instead of StMedianFilter(...) and within the MA part you can change the adaptive Temporalsoften(...) line to a "static" like as shown above Temporalsoften(2,5,8,15,2) and by this you still maintain the advantage of adaptive blurring in case of fast moving scenes.

Quote:

And I do have the PicVideo MJPEG, but in my opinion the HuffyUV capture looked a lot better even with MJPEG set at 20.

The you got "very!" good eyes and you shurely will see that MA will end up in still noise if not tweaked

BTW: You should capture THE EXACTLY same broadcasted source to do comparisons as EVERY broadcasting comes to you in a diff. quality, even if its digital! Cause its not only the "way" to your Harddisk or the signal influence of you wire or PC .... many times the stream which is originally stored and provided by the station is already "full of noise crap"!

Do have a look at this "interlacing" bible!

http://home.arcor.de/scharfis_brain/...esInterlacing/

This is like "100fps.com" but with an afterburner according to specific (but still clear!) explanations.
Use Google or something els for translation.

I never saw such a understandable and clear guide accorded to interlacing ways/techniques before.

[e1miran]

Quote:

Originally Posted by incredible

The you got "very!" good eyes and you shurely will see that MA will end up in still noise if not tweaked

BTW: You should capture THE EXACTLY same broadcasted source to do comparisons as EVERY broadcasting comes to you in a diff. quality, even if its digital! Cause its not only the "way" to your Harddisk or the signal influence of you wire or PC .... many times the stream which is originally stored and provided by the station is already "full of noise crap"!

I do have very discerning eyes. I'm not a quality freak (I don't think so), but I will try to squeeze out as much quality as possible.

I'll try some more tests, but that was my first impression when I compared HuffyUV vs. PicVideo's MJPEG. I was able to see a difference in the opening scene --the apparition of the Disney Logo on a black background. The Disney Logo was clearer/cleaner on the HuffyUV capture.

They were both the same exact source material. I was capping from my DishNetwork PVR (similar to TiVo).

[incredible]

You're right as HuffYUV does output as best as its Lossless. Every DCT based (and so also mjpg) compression will end up in quality loss, but Q=19 most times gave me a very good output, so I do come to the conlcusion that you got a PERFECT provided signal/stream by your TV station

Congratulations! (not ironic)

Sunday I captured "007 the world is not enough" .... I deleted it afterwars cause that quality was "crap"!! ... blurry, FULL of noise and so on ... and the reason wasn't my setup/Hardware as a broadcasting afterwards on the same channel came out very brilliant and sharp.

[e1miran]

Quote:

Originally Posted by incredible

I do come to the conlcusion that you got a PERFECT provided signal/stream by your TV station

Congratulations! (not ironic)

Some channels are cleaner than others-usually not by great amounts, I can attest to that. This particularly channel was TOON Disney. It figures that the channel that comes in the clearest is the one that my non-discerning 4 yr old watches most :LOL:

[kwag]

@e1miran,

Read here: http://www.kvcd.net/forum/viewtopic.php?t=8243 for the truth about the resolution of Dish Network

-kwag

[e1miran]

Quote:

Originally Posted by kwag

@e1miran,

Read here: http://www.kvcd.net/forum/viewtopic.php?t=8243 for the truth about the resolution of Dish Network

-kwag

I'm at work, so I'll read that article that you referrenced in the other post later tonight. But from the gist of you your other thread, are you saying that some networks take liberties with the quality of their encodes?

I can see how this is true. Most of my cable networks are really clear, but the channels from my local ABC, NBC, CBS, and FOX affiliates sometimes look like crap--especially "Live" programs, like American Idol.