how to slow down interlaced Ntsc recording of Pal ?

[spiritgumm]

Have an interlaced Ntsc dvdr at 29.97 fps. The movie was likely broadcast as Pal, as it's slightly shorter and pitch is high. There are several dvd recorder menus within the recording; the last recording was from Pal dvd (in dvd player) to Ntsc dvd recorder. Sample included below.
I'd like to get the movie back to it's correct speed (23.976), so I imagine it needs to get converted to 25 fps, then slowed down to 23.976 fps. I'd like encodes for dvd and for xvid, but I'm not overly familiar with deinterlacing.
Would this be apt for a new interlaced dvd encode? It looked very interlaced (too many lines) when I previewed it in MPC.
assumeTFF()
yadif(mode=1,order=1)
changefps(50)
assumefps(47.95, true)
SeparateFields()
SelectEvery(4,0,3)
Weave()
For the xvid, I tried a few variations (with last 3 lines omitted) but there was an occasional jump (like a missing frame).
VTS_01_3.demuxed.m2v

[manono]

TFM()
TDecimate(Mode=0,Cycle=6,CycleR=1)#return to 24.975fps
AssumeFPS(23.976)#slow to 23.976fps

You'll want to fix the black levels also.

[jagabo]

That clip was made by applying 3:2:3:2:2 pulldown to a 25 fps source. Try:

Mpeg2Source("filename.d2v")
AssumeTFF()
TFM()
TDecimate(mode=2, rate=25)
AssumeFPS(23.976)

manono beat me to it. I was also going to suggest the mode 1 method.

[spiritgumm]

thanks, guys! I thought pulldown was still progressive, so when I was checking for interlacing (added assumebff) and got the backward motion, I thought it was interlaced.
Are there pros/cons between the two TDecimate settings? Would TDecimate(mode=2, rate=25) be useful in case the pattern is every irregular? Should it's rate be 24.975, or does it not matter here?

For the black level, I followed the advice from an old thread which I hope is still applicable (http://forum.doom9.org/archive/index.php/t-70140.html) and added
Levels(0,1.0,255,0,255)
Histogram("levels")
and adjusting 1st/3rd parameters until the top graph filled horizontally, which was about Levels(38,1.0,176,0,255).

[manono]

Your levels settings are screwed. Just add a coring=false to it and then have a look. Also, to help find the proper settings, crop away all the black so just the video itself is visible, something like:

Crop(16,16,-16,-16)
Levels(38,1.0,176,0,255,Coring=False)
Histogram("levels")

Then it should become pretty obvious what you did, even if you can't tell from just looking at the video you created (although you should easily be able to tell something's dreadfully wrong just by looking at the result you created). I like rich and deep black levels so it might be as simple as:

Tweak(Sat=0,Bright=-33,Cont=1.1,Coring=False)

When done checking then I'd do the final crop and resize, making sure to apply fresh black to the sides. And with more of the video available to check, I might also make other adjustments to the gamma/brightness, and/or contrast levels. You can make similar adjustments using ColorYUV and Levels. Me, I usually use Tweak and YLevelsS.

[spiritgumm]

I initially tried Tweak before Levels, but didn't know Bright could go negative. I only tried Levels because it came up when I was searching "Black Level avisynth." Is there an generic method for getting a correct Black Level (with Tweak and/or YlevelsS), similar to the histogram advice in that old thread? I can't trust my monitor or my eyes.

[manono]

In addition to a histogram, you can also use:

ColorYUV(Analyze=True)
Limiter(Show="Luma")

Red and green are the levels below 16 and above 235, respectively. It'll also show you the max and min luma values and kind of an 'average' max and min value. It's very useful and that's what I use most of the time.

You could always calibrate your monitor, you know. Jagabo has cards he's posted before that can help with that. Or maybe your monitor came with calibration software. Also, if the DVD is intended for display on a TV set, ordinarily if it looks 'good' on a monitor, it's too contrasty or 'bright' for a TV set. Good TV luma values often look a bit 'dull' or even 'dark' on a computer monitor. This also assumes a calibrated TV set, which almost none are out of the box.

[jagabo]

This post:

https://forum.videohelp.com/threads/326496-file-in-Virtualdub-has-strange-colors-when-o...=1#post2022085

Has a DV AVI levels video that's useful for calibrating your monitor and testing your work flow.

[spiritgumm]

I have noticed the brightness difference between monitor and TV. I've often wondered when encoding for AVI file sharing, what do the majority of people use for viewing? I watch everything on a TV, but I've often brightened videos for xvid because I thought the majority watch on a monitor/laptop. Should I not assume that, and only correct for TV viewing?

Since the min/max values change alot, what do I base my decision on for values in YlevelsS? I also don't know what to do with the histogram, since trying to adjust the high/low inputs to touch the brown borders doesnt look good.

[jagabo]

Since the min/max values change alot, what do I base my decision on for values in YlevelsS?

Unless you want to adjust the video scene by scene you have to use an compromise setting for the entire movie. VHS caps are the worst in this respect because the automatic gain in VHS deck puts the black and white levels all over the place.

I also don't know what to do with the histogram, since trying to adjust the high/low inputs to touch the brown borders doesnt look good.

You should go through the movie and adjust the black level for the darkest parts of the video and the white level for the brightest parts. That will usually give you a fair compromise. If you adjust the black level by a large amount it usually helps to adjust the gamma too.

[spiritgumm]

I'll see if the gamma helps. Bright areas at different parts of the movie dont peak the same.
I downloaded the test pattern. The monitor whites gibe for TV (=>235), but black range is larger- up to 25 or 30. Does this correspond with the generally darker picture on a monitor versus TV?

[jagabo]

I'll see if the gamma helps. Bright areas at different parts of the movie dont peak the same.

That's why you have to search for the brightest sections and adjust for that.

I downloaded the test pattern. The monitor whites gibe for TV (=>235), but black range is larger- up to 25 or 30. Does this correspond with the generally darker picture on a monitor versus TV?

Many monitors (and TVs) have trouble down in the darkest bars. Optimally you want to be able to see the 20 and 25 bars but don't fret if you can't. You can try zooming in to see it a little better (with less bright stuff on the screen it's easier to differentiate the black bars).

[spiritgumm]

But the test pattern (for me) is it to adjust my monitor to display as TV so I can make better video level adjustments.

Inre ColorYUV(Analyze=True), I'm using the highest value of the Min and Max for my YlevelsS Input Low and High, right?

[jagabo]

But the test pattern (for me) is it to adjust my monitor to display as TV so I can make better video level adjustments.

You can use it to adjust your display and to test your workflow while editing -- to verify that levels aren't being messed up somewhere in the process. By the way, I would first use an RGB chart on the desktop to adjust the monitor (with the Desktop's proc amp set to neutral). Then use the video file to adjust your graphics card's video proc amp settings (with a player set to neutral).

Inre ColorYUV(Analyze=True), I'm using the highest value of the Min and Max for my YlevelsS Input Low and High, right?

I wouldn't do that. Especially if you don't crop away the borders. The borders are often darker than the picture content. And even if you do crop away the borders there are often outliers that are especially dark or bright (oversharpening halos, for example). You don't care if a few pixels here and there get truncated.

A trick when using Histogram:

TurnRight()
Histogram()
TurnLeft()

That will give you a horizontal graph at the top rather than a vertical graph at the right.

[spiritgumm]

I cropped the borders for the xvid and was going to keep those same YlevelsS values for the mpeg2 encode (which will have borders). So how am I to interpret and use ColorYUV(analyze) ? I'm already encoding the xvid, but I'll try that histogram tip when I'm done.

[manono]

The way I use those 2 lines of code is by making sure there aren't any large blobs of red or green. It's OK to have scattered dots here and there, but no large masses of color. And most of the films with which I work start with large blobs. They've been converted to RGB somewhere in the workflow, everything's completely messed up, and often it can't be entirely fixed as the damage has been done already. One thing you can do to train yourself in the use of histograms or those 2 lines of code I showed you before is to study how the pros do it. Whenever you open a good Hollywood produced DVD in your computer, use scripts with a histogram or the 2 lines and study how they do it. Of course, sometimes in the quest for a certain 'look' they'll mess with the brightness/contrast/levels/saturation or they'll intentionally add grain, but those are usually pretty easy to tell. The Hollywood DVD production companies are the best in the business (with the possible exception of Criterion, if you come across any of their work) and you can learn a lot by studying how they do it.

Sometimes 2 or more different sources are used to produce an old movie DVD and they might have different brightness/contrast values for the different parts. If so, you can trim off sections and apply your filters differently to different parts. Otherwise, unless you're willing to do a ton of trims (and I've done it before), about all you can do is try and find a happy medium, as jagabo mentioned earlier.

[spiritgumm]

That's a good idea - I have a couple Criterions, both color and b&w.
Since this movie is b&w, I'm not seeing any green/red blobs. If I'm suppose to, then something is wrong on my end.

[jagabo]

Since this movie is b&w, I'm not seeing any green/red blobs. If I'm suppose to, then something is wrong on my end.

It doesn't matter that the video is black and white. Limiter(Show="Luma") replaces pixels that are below IRE 0 with red, above IRE 235 with green for easy identification. Your video after too much contrast stretch:


and after a more reasonable contrast stretch (for this frame):


the second image with VideoScope():

[manono]

Right, your problem is that all the black in the video isn't black but grey. You want to make it blacker. There isn't any red to be seen. Take down the black levels until you begin to see red. In my script for your sample earlier, my filters brought it down until red dots were begun to be seen. Because Tweak's brightness also messes with the contrast, you'll have to raise it a bit to compensate. I made it how I like it, though. You may or may not want a similar 'look'.

I think jagabo works mostly with ColorYUV and the histogram. Just learn to work with whatever suits you.

[jagabo]

I think jagabo works mostly with ColorYUV and the histogram. Just learn to work with whatever suits you.

I usually use ColorYUV() and VideoScope() for levels adjustment. But that's just a personal preference.

[spiritgumm]

Ah, the problem was on my end. I didn't carry over Limiter when I was doing further testing. I've already started my encode using the manono's Tweak script. I'll recheck the original video with Limiter when it's done so I can really digest what you're saying.