Amazing discovery in capture resolutions (IMHO)

[SatStorm]

Real Interesting...
Indeed, my Hauppauge, older card sharps well on values between 50 - 55, while my Kworld, newer card, is capable to sharp better on values between 65-75. My tests also show me that for kworld, the 81 value is oversharped.

It is great for cartoon that way

Anyway, I expect objervations and discussion about this with other users.
I believe that if that tweak drivers had some kind of manual, we wouldn't have to test ourselfs to understand the possibilities. Of course, that means less fun also....

[tito13kfm]

this post has enlightened me on so many issues and has also raised many questions in my mind.

satstorm mentions:

One interest thing in case we wish to have the best possible aspect:
Grabb at 344 x 576/480 and expand to 352 x 576/480. With the sharpen filter activated manually, the picture is much better that way and with the correct aspect!

This would be fine for the cards that capture 688 lines natively.

I'm assuming this would not apply to the older BT cards (using BTWinCap driver) which capture 712 lines natively, where half would be 356x480. So in this instance I would capture at 356x480, crop 2 pixels from each side and encode normally. Am I correct in my thinking, or have the last 4 pages of information left me dumber?

[SatStorm]

bt8xx cards don't capture boarders (if those are detected), so the aspect is almost alway a little distort.
344 x 576 expand to 352 x 576, adds some boarders which in practice restore the aspect of the transmission (when on a transmission an overscan exist).

From satellite DVB channels, there are always some borders left and right. My bt8xx cards don't seems to grab those: So that alternative way, someone who wish aspect about the same at what he sees on TV, might be a practical solution.

My presentations are not based on maths, but on objervation and practical use. I leave maths to those who can deal with them. I'm not one of them...

[tito13kfm]

bt8xx cards don't capture boarders (if those are detected), so the aspect is almost alway a little distort.
344 x 576 expand to 352 x 576, adds some boarders which in practice restore the aspect of the transmission (when on a transmission an overscan exist).

I've noticed after some playing around with my card that it captures a left border but no right border.. so I guess i'll cap at 344x480 and expand all 8 pixels on the right to try and balance it out.. or I could just say screw trying to keep perfect AR and cap at 352x480 like I have been doing, only override the sharpness setting.

[trevlac]

@tito

Try This. And see if it helps. Everything depends upon what the driver is doing. Talking about borders can be misleading. It depends upon the driver and the source, if you talk about borders.

I made that page a while ago. Best to try to understand what is going on, than to blindly follow my advice.

Trev

[deadpac]

Yeap, exactly, you confirm it...

Just compare 352 x 480 sharpen at 51 with 368 x 480 on auto....
About (if not...) the same picture.
Set now the sharpen filter to the highest possible value (81) and prepare for great suprises!

I did just that and here is what I got. Well not exactly, 51 seemed way to soft and so I upped it to the max value 81 and I still came up with results opposite to what you suggest.

Maybe I'm missing something here but I think my card's (OLD PCTV Pro w/Btwincap drivers) is going to want to disagree with you. Even in fmctm1sw's post the text in the 352x480 and 368x480 captures with the value of 50 shows a distinct difference in sharpness. The "Survivor: All-Stars" text is much more pleasing to my eyes in the 368x480-50 picture than the slight-hazy-blurry 352x480-50 cap.

So based on several tests I did (H:352,368 S:49,50,55,81)and with some help from fmctm1sw's post, it seems like 368x480 does have a sharper picture than 352x480 with the same sharpness setting; be it 49, 50, 55, 81, or auto.



The text to me on the left looks much sharper than the text on the right.

[SatStorm]

Sure it does...
Afterall, not all bt8xx(x) cards are the same right?
Seems that yours are even older (or based on an older cheap) than my Hauppauge, and seems that the sharpness is less.

Also, I didn't say that just uncheaking 352 x 576 gave me sharpness equal to 358 x 576. I had to rise the value to make it look like the 368 x 576 capture!
And it is probably not a natural sharpness, but it does the job for my needs.

For my kworld bt878a card, setting sharpness somewhere between 65 - 70 give me results I really like.
For my Hauppauge card, it is somewhere between 55 - 60. Lower blures, higher sharps unnatural

It is an interesting hobby, isn't it?

Also, keep in mind that I'm PAL. My objervations are totally different than fmctm1sw which is NTSC. It is nature I think, the maths are different. In the matter of fact, I'm along in this: Still no PAL user testify his results about this setting.....

Hey, PAL users with bt8xx(x) cards! Do some tests, please! This gonna make your captures better!

And, I repait: That happens when we use universal third party drivers and we don't have any documention / instructions for them.

[trevlac]

@deadpac

The difference (in my NTSC experience) between 368 and 352 is the vertical. You can see this in your pic in the word "Power". 368 does not have a vertical filter applied. 352 does. Even though you do not resize on the vertical, the filter is turned on for 1/2 D1. It blurs on the vertical. In my opinion, VHS vertical is even better.

This was my original point. I have not been following all of Stat's results. I assume he is happy with what he gets. I'm not sure what happens on all of the variations of cards and drivers. It is easy to find out, I posted a link above that allows you to look at the exact options set for the chip. One can of course just do some test caps also. I'd recommend a resolution chart. AVIA or Video Essentials are disc with these, or you can make one from free patterns on the net. The problem jumps out on a thin horizontal line.

Cheers

[DereX888]

Hmm, after reading this thread for about a month? (IMO one of the most interesting threads nevertheless) we are back to square one lol! As usual the more people do the tests, the more confusing results are

trevlac - congrats for spotting this nuisance
however it all is dependant on a hardware version more than the driver and its setting it seems after all ? (as in deadpac's samples)

[fmctm1sw]

@trevlac - that bttool is very interesting. I capture at a width of 712 pixels no matter what it seems. I just have to wonder what method of "shrinking" my card/driver apply to my picture and if I would be better off capturing at 712 and resizing myself.

Edit: And I almost forgot... When you guys mention "padding" and "adding pixels," how exactly are you doing that? Avisynth? Virtualdub? I use both...

[FulciLives]

Edit: And I almost forgot... When you guys mention "padding" and "adding pixels," how exactly are you doing that? Avisynth? Virtualdub? I use both...

I always use AviSynth

I use VirtualDubMod to edit then I use the SCRIPT EDITING function found in VirtualDubMod to import my edits into an AviSynth AVS script.

I then load the AviSynth AVS script into CINEMA CRAFT ENCODER

I use PICVideo MJPEG which uses YUV or YUY2 or whatever and this is the colorspace that CCE works in so by using AviSynth and CCE I never have to worry about any YUV to RGB conversions.

- John "FulciLives" Coleman

[trevlac]

@fmctm1sw

Yep... all card/driver combos seem to capture at 1 size. (Not to confuse matters, 712 is the size given in 13.5MHz dvd pixels. Giving the size as a portion of the scan line would be more accurate, but no one would know what that means .)

On the resize thing, I have been arguing with some heavy weights a doom9 and the virtual dub forum. about this very question. I'd say the BT resize is as good as any post capture resize. So letting the BT do it, or doing it after does not matter.

You can also crop as you cap, if you use virtualVCR. There is a built in crop filter. Cap at 712 and crop to 704 ... presto.


@DereX888

Maybe it is all dependant ... but if you have a BT8x8 chip and you use BTtool, and you look at the tables in the back of the spec document, you can determine what is going on. But at some point is may not be worth the time.


@FulciLives

Ya know, I think that proper color makes more of a difference to a good picture than more resolution at some point. But now you're getting into a whole nother confusing story ...

[vhelp]

please excuse me for barging in here like this..

but, can anyone tell me how to write the .AVS script that allows you to include
text in the video pics, much like the ones that deadpac used in his pics
above ??

Thanks,
-vhelp 2542

[deadpac]

vhelp,

I just use the built in Subtitle Filter

AVISynth[/url] 2.5 Documentation]
Subtitle

Subtitle(clip, string text, int "x", int "y", int "first_frame", int "last_frame", string "font", int "size", int "text_color", int "halo_color" int "align" int "spc")
Subtitle(clip, string "text")

The Subtitle filter adds a single line of anti-aliased text to a range of frames. If you want more than one subtitle, or a subtitle that takes more than one line, you have to chain several Subtitle filters together. This filter is obviously unsuitable for anything more than occasional use. I use it occasionally.

The text color and halo color should be given as hexadecimal RGB values, as in HTML--except that they start with $ instead of #. (I was already using # to introduce a comment.)

The alternate, short form is useful when you don't really care what the subtitle looks like as long as you can see it--for example, when you're using StackVertical and its ilk to display several versions of a frame at once, and you want to label them to remember which is which.

This filter is used internally by Avisynth for the Version command and for reporting error messages, and the subtitling apparatus is also used by ShowFrameNumber.

align=4 (left,baseline) or 5 (center,baseline) if x=-1; (align position represented by 1-9 on numeric keypad);

spc=0 (no expansion / contraction of text spacing)

x=8 if align left (1,4,7); x=center if align center(2,5,8 ); x=right-8 if align right(3,6,9);

y=0 if align top (7,8,9); y=size if align baseline(4,5,6); y=bottom if align bottom(1,2,3);

specifying -1 for x or y results in center coordinate for that axis (previously only for x and was undocumented)

If you just want to label videos that are being compared just a simple...

AVISource("OriginalAVI.avi").Subtitle("Original Video")

The text will always go in the upper left corner of the video it is applied to and appear throught the video with just a simple subtitle("some text")

If you want to get a little more complicated just define two (or more) video variables one for the original video and another video that has been filtered and just play with either StackVertical() or StackHorizontal() or both.

I've saved 3 subtitle scripts that have their font size and positioning based on the frame size I want them applied too. One for 352x240, 352x480, and one for 720x480. Paired with BlankClip and you have a cheap way of making information videos that you can either apply to the beginning or end of a video.

It would probably be a lot easier to do something like this in Premiere rather than AVISynth. But for the quick and dirty text jobs I just stick to AVS 8)