Worst DVD transfers ever (please sticky)

[Katie Boundary]

This thread exists to serve as a warning for everyone who might wish to extract footage from certain problematic DVDs. PLEASE offer contributions to this list so that as many people as possible can be warned.

FILM content encoded as hard NTSC

Sometimes, a show that was shot or drawn at 24 progressive frames per second, with no 30 FPS or natively interlaced content whatsoever, will be encoded as hard NTSC because some DVD authors are just dicks like that. The best way to deal with this kind of content is to use uncomb(), and then look at the resulting stream frame by frame in Virtualdub. If you notice nothing out of the ordinary except that every fourth frame is duplicated, then go ahead and add Decimate(); however, if there are still a few interlaced frames because an incompetent telecine job left orphaned fields, you'll need to scrap Uncomb and replace it with Bob(), followed by either Selecteven() or Selectodd(), followed by decimate(). If you're downscaling to 288p or less, you can use Separatefields instead of Bob() to save a few CPU cycles.

Examples: Transformers (the original series), later seasons of South Park, a few random isolated episodes of Kim Possible, and the "Platinum"/Thinpack" releases of Neon Genesis Evangelion.

PAL/SECAM content encoded as hard NTSC
I first learned about this phenomenon when playing around with ripped War Planets DVDs. Solution: Uncomb() followed by Decimate(6). Or just get the Region 2 discs like a civilized person. In the extremely specific case of shows made by Mainframe Entertainment, which actually duplicated 199 out of every 1000 frames (or whatever) instead of slowing down by 0.1% and then duplicating 1 out of every 5, you can use Chidragon's plugin to avoid accidentally deleting every thousandth frame.

Examples: War Planets, Reboot III, possibly other Mainframe shows. Given that Code Lyoko is a French show that became pretty popular in the USA, I'd bet good money that the American distribution of Code Lyoko falls into this category as well. Can anyone check?

Mixed-framerate content
These shows shot live-action footage at 24 progressive frames per second, rendered their special effects at 30 progressive frames per second, and interlaced them together. The best solution is to use Uncomb(). After this, you'll have to decide whether you'll tolerate the duplicated frames in the live-action content or use Decimate() and tolerate the skipping in the effects footage. Preserving the mixed framerate is also possible, and even preferable for certain extremely specific purposes, but the methods are elaborate and beyond the purpose of this guide.

Examples: Babylon 5, The Secret World of Alex Mack, at least 3 of the Star Treks (TNG, DS9, Voyager)

Natively 60i content

While almost no shows have been done entirely in native 60i since the end of the Cold War, many of them at least include short special effects that are rendered in this way. Fades, especially in opening credits or title sequences, are also frequently done in 60i. The ONLY way to clean up this footage is to use Bob(); Selecteven() and Selectodd() are optional depending on whether you want 30p or 60p. If you're downscaling to 288p or less, and you plan on using Selecteven() or Selectodd(), then you can use Separatefields() instead of Bob() to save a few CPU cycles.

Examples: Numerous effects on Gene Roddenberry's Andromeda (shakeycam, entering or exiting slipstream, the "banding" on Rommie's hologram), flashbacks on Birds of Prey, C5 Corridor sequences in Where on Earth is Carmen Sandiego, anything that was originally shot "live" (like that one episode of ER in which George Clooney waved at the camera), most older sitcoms.

Badly captured content from VHS or Beta

The second-ugliest of all kinds of footage. Using separatefields() or similar reveals that this is not just effectively 60i, but that each pseudo-field is made up of two of the original film frames blended together, sometimes unevenly from top to bottom. Do not use footage like this unless you absolutely must; if you must, then your only safe bet is Bob(). Selecteven() and Selectodd() are optional depending on whether you want 30p or 60p. If you're downscaling to 288p or less, and you plan on using Selecteven() or Selectodd(), then you can use Separatefields() instead of Bob() to save a few CPU cycles.

Examples: numerous episodes of Captain Planet, the original domestic releases of Neon Genesis Evangelion

Natively NTSC content converted to PAL

When Film content gets converted to PAL, it typically just gets sped up by 4% and no one notices. However, if something was specifically designed for NTSC and thus contains 30p or 60i content, then the conversion to PAL will absolutely murder it. Expect most if not all of your frames to be blended. Solution? Buy the region 1 DVDs and use those instead. Then you'll have the same nightmares as the rest of us. In fact, this is also what you should do for anything that was originally done at 24 fps.

The interlacing from hell

I have no bloody idea how content like this even comes to exist. It's as if the original footage gets interlaced, then resized, then interlaced again. Even Separatefields cannot help you here. Your only hope of extracting clean footage is to make a deal with Satan.

Examples: the first few seasons of South Park, parts of the opening title sequence of Batman Beyond.

[huhn]

nearly ever anime DVD is hard telecine.

it is that worse that there was/is a list listing the animes that are soft telecine.

and why should i use uncomb if i have access to TIVTC or animeIVTC?

[hello_hello]

I'm not sure film content encoded with hard pulldown qualifies for a "worst DVD transfers ever" list, given it's not particularly uncommon and should be reversible. Is it an incompetent telecine job that leaves orphaned fields or is it the editing of hard telecined material?

I live in PAL-Land so I'm far from an NTSC expert but unless I'm certain an NTSC DVD is 100% soft telecine I'd decode it with the pulldown flags being honoured and treat it as hard telecine anyway. Options for restoring hard telecined content also include using something better at field matching than UnComb() or field matching/decimation that includes the de-interlacing of combed frames. SeparateFields().SelectEven() halves the vertical resolution.

Vinverse (http://avisynth.nl/index.php/Vinverse) also seems popular for cleaning up residual combing.

For "mixed-framerate content" I thought standard procedure is to convert to a common frame rate using frame blending rather than simply dropping or duplicating frames. It's not ideal either, but it's more fluid. Other options these days might include converting to a 60fps or 120fps constant frame rate. 60fps could at least use a 3:2 progressive pattern for the "film" parts (which is what you get when watching 24fps progressive on a monitor refreshing at 60Hz anyway) rather than repeating every fourth frame at 24fps. The former is far less obvious to me. I can't say I've investigated either possibility when converting DVDs, but I imagine it's do-able. You could also treat mixed frame rate content as variable frame rate.

"60i content"? Isn't that the definition of NTSC interlaced video? There's de-interlacers for that sort of thing, rather than throwing away half the fields. 60i would qualify as the non-film part of "mixed-framerate content", but it's interlaced rather than progressive. After de-interlacing you're left with a mixture of 30fps progressive and "film" with either soft or hard telecine.

Mostly I'd let TIVTC (http://avisynth.nl/index.php/TIVTC) sort it out for me.

There's other types of pulldown, such as for converting PAL to NTSC. I don't know how often they'd appear on DVD but there's a documented case for Bluray here (http://forum.doom9.org/showthread.php?p=1754249#post1754249).

"PAL/SECAM content encoded as hard NTSC". I'm not sure what that's describing exactly, but in a perfect world you'd want to restore the original 25fps PAL rather than treat it as NTSC with 3:2 pulldown applied. Is there field blending involved? It sounds like it might be a job for SRestore() (http://avisynth.nl/index.php/Srestore), which also does a pretty good job of repairing hard telecined NTSC that was incorrectly deinterlaced for 29.970fps progressive rather than IVTC'd for 23.976fps.

"Captured content from VHS or Beta". They both should be 29.970fps interlaced for NTSC and captured that way without a need to blend anything. Something horrible could have been inflicted on the video before it was transferred to VHS or Beta though, or at some other stage. If you use Separatefields() on it without SelectEven() or SelectOdd() following, wouldn't you end up with a bobbing effect?

The reason for "interlacing from hell" is a mystery to me. Often it's progressive video that appears to have been interlaced in a previous life and resized before it was de-interlaced. If it's progressive, throwing away every second scanline with separatefields().selecteven() can reduce the effect a little if you follow it with nnedi3 to interpolate them (rather than simply resizing the remaining fields to the original resolution). ie nnedi3(dh=true).

[kuchikirukia]

This has to be a troll thread.

[Katie Boundary]

and why should i use uncomb if i have access to TIVTC or animeIVTC?

I have no idea because I've never used either of those. However, because every interlaced video source has its own quirks, IVTC or deinterlacing methods will need to be customized and tweaked for each source, so you'll generally need to use uncomb anyway. "Smart" filters are no match for a human when it comes to dealing with unexpected problems.

Uncomb is a particularly useful diagnostic tool for figuring out which video sources can be IVTCed using conventional means and which ones require a more aggressive approach.

I'm not sure film content encoded with hard pulldown qualifies for a "worst DVD transfers ever" list, given it's not particularly uncommon and should be reversible. Is it an incompetent telecine job that leaves orphaned fields or is it the editing of hard telecined material?

I'm not sure. I just know that this content exists and SOME of it must be handled carefully to avoid interlaced or blended frames. I agree that the stuff that IVTCs perfectly doesn't really qualify for a worst-ever list, but it's impossible to tell from the nasty stuff without doing an Uncomb diagnostic.

unless I'm certain an NTSC DVD is 100% soft telecine I'd decode it with the pulldown flags being honoured and treat it as hard telecine anyway.

It's not a guessing game. DGIndex will tell you exactly what kind of content you're dealing with as the project file is being generated.

SeparateFields().SelectEven() halves the vertical resolution.

Wow, if only AVIsynth had a resize filter that could fix this problem...

For "mixed-framerate content" I thought standard procedure is to convert to a common frame rate using frame blending rather than simply dropping or duplicating frames. It's not ideal either, but it's more fluid.

Ugh, hell no. I don't consider blended frames to be any better than interlaced frames (both waste bitrate by screwing with motion-search), and my inferior American eyes are already accustomed to slight judder :)

Other options these days might include converting to a 60fps or 120fps constant frame rate. 60fps could at least use a 3:2 progressive pattern for the "film" parts (which is what you get when watching 24fps progressive on a monitor refreshing at 60Hz anyway) rather than repeating every fourth frame at 24fps. The former is far less obvious to me. I can't say I've investigated either possibility when converting DVDs, but I imagine it's do-able.

That option isn't very useful for converting to VCD-compliant MPEG-1, or making anime music videos, or much of anything.

You could also treat mixed frame rate content as variable frame rate.

If you wanted to store everything in some obscure format like Oggmux or Matroska that only 0.1% of people have ever heard of and would be capable of watching, sure.

"60i content"? Isn't that the definition of NTSC interlaced video?

Not quite. I'm talking about content where 60 unique frames per second are generated, one field from each frame is discarded, and the surviving fields are then weaved together. I provided some fairly specific examples if you'd like to know what I'm talking about, though you'd have to get the Region 1 versions of the DVDs.

"PAL/SECAM content encoded as hard NTSC". I'm not sure what that's describing exactly

Step 1: find PAL/SECAM content.

Step 2: duplicate every 5th frame instead of every 4th.

Step 3: Interlace.

Profit?

but in a perfect world you'd want to restore the original 25fps PAL rather than treat it as NTSC with 3:2 pulldown applied.

Sorry, I wrote "Decimate(5)" when I should have written "Decimate(6)". The error has been corrected.

"Captured content from VHS or Beta". They both should be 29.970fps interlaced for NTSC and captured that way without a need to blend anything. Something horrible could have been inflicted on the video before it was transferred to VHS or Beta though, or at some other stage.

Ive seen things... horrible things... things I won't describe here...

If you use Separatefields() on it without SelectEven() or SelectOdd() following, wouldn't you end up with a bobbing effect?

Yes. Some people might find it tolerable at 60 Hz. I think there is also an AVIsynth filter that attempts to correct this bobbing.

throwing away every second scanline with separatefields().selecteven() can reduce the effect a little if you follow it with nnedi3

Tell me more about this "nnedi3"...

[kuchikirukia]

This has to be a troll thread.

[Katie Boundary]

This has to be a troll thread.

Your mom is a troll thread.

[hello_hello]

I have no idea because I've never used either of those. However, because every interlaced video source has its own quirks, IVTC or deinterlacing methods will need to be customized and tweaked for each source, so you'll generally need to use uncomb anyway. "Smart" filters are no match for a human when it comes to dealing with unexpected problems.
Uncomb is a particularly useful diagnostic tool for figuring out which video sources can be IVTCed using conventional means and which ones require a more aggressive approach.

Uncomb() is simply a field matching filter, and by it's own admission, not particularly clever. Do you think maybe there's a happy balance somewhere between using an automated tool and checking the result with your eyes, and could it be possible a "smart filter" might shift the balance a little rather than force you to completely rely on it?
You never know, it might even be smart enough not to have an option that throws away half the fields and halves the vertical picture detail, but simply de-interlaces the occasional remaining combed frame instead.

What's the advantage of Uncomb() over Decomb's own field matching, or over using TFM() from the TIVTC plugin?

I'm not sure. I just know that this content exists and SOME of it must be handled carefully to avoid interlaced or blended frames. I agree that the stuff that IVTCs perfectly doesn't really qualify for a worst-ever list, but it's impossible to tell from the nasty stuff without doing an Uncomb diagnostic.

Well I'd argue that checking the result of an automatic filter also qualifies as doing a diagnostic, but most people would check the source before deciding how it should be handled and which tool to use, and even then they'd check the result. Throwing your hands up in the air and tossing half the fields away after detecting an occasional orphaned field would probably be most people's definition of not handling a source carefully.

It's not a guessing game. DGIndex will tell you exactly what kind of content you're dealing with as the project file is being generated.

I was referring to not using "force film".

Wow, if only AVIsynth had a resize filter that could fix this problem...

Really, you didn't get that I was referring to resolution in respect to picture detail, not picture size? Really, you didn't get that?

Ugh, hell no. I don't consider blended frames to be any better than interlaced frames (both waste bitrate by screwing with motion-search), and my inferior American eyes are already accustomed to slight judder :)

There's nothing even remotely "slight" about the judder produced when dropping or repeating frames to convert 23.976 and 29.970 to a common frame rate. It looks terrible.

That option isn't very useful for converting to VCD-compliant MPEG-1, or making anime music videos, or much of anything.

Well not to be picky, but your opening post clearly implies it doesn't matter if the video's not very useful for converting to VCD-compliant MPEG-1, or making anime music videos, or much of anything. Why the change of heart?
"The ONLY way to clean up this footage is to use Separatefields(); selecteven() and Selectodd() are optional depending on whether you want 30p or 60p."

If you wanted to store everything in some obscure format like Oggmux or Matroska that only 0.1% of people have ever heard of and would be capable of watching, sure.

Seriously though..... it's 2016. My 82 year old mother has a hard drive connected to her TV's built in media player. Almost everything plays MKV these days and if it doesn't it'll play MP4, which it may surprise you to learn, supports variable frame rate too.
Support for High Profile Level 4.1 AVC is the norm these days, and support for higher levels is becoming more common now you can buy discs that look like VCDs but hold 4k video. High Profile, Level 4.1 supports 60fps up to (and including) 720p.

De-interlacing DVDs to 60fps (or 50fps) progressive is fairly standard now because hardware players generally have no problem with it, it looks much smoother than 30fps (or 25fps) and when you play an interlaced DVD, that's generally how it's being de-interlaced on playback anyway.

Not quite. I'm talking about content where 60 unique frames per second are generated, one field from each frame is discarded, and the surviving fields are then weaved together. I provided some fairly specific examples if you'd like to know what I'm talking about, though you'd have to get the Region 1 versions of the DVDs.

Well I'll confess that's something I don't think I've seen before.
I'll confess I'd also put it in the "sounds unlikely" category, purely because I'm not sure why 60p content that's not much good for anything was being created so often in the first place. I've been wrong before, but do you have a sample? It might be handy to include some samples in a "Worst DVD transfers ever" thread aimed at sticky status so others can see for themselves.


Step 1: find PAL/SECAM content.

Step 2: duplicate every 5th frame instead of every 4th.

Step 3: Interlace.

Profit?

Sorry, I wrote "Decimate(5)" when I should have written "Decimate(6)". The error has been corrected.

Why would you just repeat a frame? Fields are repeated, otherwise it's not really pulldown in the traditional sense, it's just a repeated frame. I assume you realise for NTSC 3:2 pulldown every fourth frame isn't duplicated either, and a duplicate frame doesn't actually exist. I assume you just weren't thinking clearly when you wrote that.
That aside, would you explain the math used to show 25fps can be spread over 29.970fps by simply repeating one frame in six, or is that a "close enough" method? No orphaned field possibilities there? A more aggressive approach not required as it might be for NTSC pulldown?

Ive seen things... horrible things... things I won't describe here...

I take that to mean your assumption probably isn't correct.

Yes. Some people might find it tolerable at 60 Hz. I think there is also an AVIsynth filter that attempts to correct this bobbing.

Even if at 60Hz it's not very useful for converting to VCD-compliant MPEG-1? ;)
"Some people might" and "I think there's also an Avisynth filter" doesn't seem like very "complete" advice.

[ChiDragon]

If you've never seen OP's posting history, they have a pretty unorthodox idea of video. One might say that every thread is a troll thread.

That aside, would you explain the math used to show 25fps can be spread over 29.970fps by simply repeating one frame in six, or is that a "close enough" method? No orphaned field possibilities there? A more aggressive approach not required as it might be for NTSC pulldown?
I don't have any War Planets DVDs, but I do have the ReBoot Season 3 DVDs released by ADV that OP referred to. Not only is frame repetition obviously the incorrect diagnosis, the suggested Decimate(6) produces the wrong frame rate (24.975 fps).

The actual pulldown used by Mainframe Entertainment was direct from 25.000 -> 29.970 without any slowdown. They used a super-long pattern to accomplish this (1200 frames with 199 duplicate fields). The reason I know this is that I manually reverted each episode to 25.000 using a plugin that I wrote, and far too much time. In some episodes, there were random pattern breaks as well: my archived Override file shows that I fixed 96 occurrences in one ep. Why not just use TIVTC, etc.? The auto-decimate filters don't play well with such long cycles, and I'm apparently insane enough to put this much effort into a children's cartoon I enjoy.

1200-199 = 1001

1001/1200 is the simplest form fraction of the division 25/(30000/1001) = 25025/30000 so a shorter cycle for 25.000 is impossible.

[hello_hello]

Not quite. I'm talking about content where 60 unique frames per second are generated, one field from each frame is discarded, and the surviving fields are then weaved together. I provided some fairly specific examples if you'd like to know what I'm talking about, though you'd have to get the Region 1 versions of the DVDs.

Thinking about it again, that still seems to be standard interlaced NTSC you're describing. It's 59.940 fields per second.
What you're describing there is still 59.940 fields per second, obtained by first creating full frames, then removing half of each frame later.
Assuming the second method is even likely, I'm not sure how the end result would be any different, or how you'd tell which method was used.

[hello_hello]

I don't have any War Planets DVDs, but I do have the ReBoot Season 3 DVDs released by ADV that OP referred to. Not only is frame repetition obviously the incorrect diagnosis, the suggested Decimate(6) produces the wrong frame rate (24.975 fps).

I'm sure I've come across something similar a couple of times. I can't seem to recall what they were exactly, although they wouldn't have been animation.

I assumed, incorrectly at the time, that 25 to 29.970 conversions would always be done by first slowing PAL down to 24.975. It makes sense for the same reason film is first slowed down to 23.976. Going to directly from 24 to 29.970 would probably be just as convoluted, although maybe a complicated pulldown pattern for converting animation is no big deal. I don't know.

The Decomb filter suggests using Telecide(guide=3) for field matching that type of video and mentions a 3:2:3:2:2 pulldown pattern. 12/10x25=30.
Not that I'm trying to imply you're wrong. While I was consulting with the internet I rediscovered FDecimate (http://avisynth.org.ru/docs/english/externalfilters/fdecimate.htm), and it mentions both methods.

"Another example is 25 fps to 60 fps conversion, which uses a 3232232322... pattern. Converting from 60 fps to the broadcast rate of 59.94 fps can be achieved by either slowing down the video and adjusting the audio to match it, or by excluding one of the duplicates every 1000 frames."

I think that means if I incorrectly used something like TDecimate(cycle=6, cycleR=1) I probably didn't do too much damage. An extra frame removed from each group of 1200, which is an extra frame removed roughly every 40 seconds, assuming no random pattern breaks and if I've worked that out correctly. Still, better to get it right. The plugin to which you referred.... is that available for public consumption? :)

[Katie Boundary]

Uncomb() is simply a field matching filter, and by it's own admission, not particularly clever. Do you think maybe there's a happy balance somewhere between using an automated tool and checking the result with your eyes, and could it be possible a "smart filter" might shift the balance a little rather than force you to completely rely on it?

Uncomb's lack of cleverness is exactly what makes it such a great diagnostic tool.

You never know, it might even be smart enough not to have an option that throws away half the fields and halves the vertical picture detail, but simply de-interlaces the occasional remaining combed frame instead.

If this is achieved by any means other than putting humpty dumpty back together again from his original fields, then it's not an improvement over just scrapping half of the fields.

What's the advantage of Uncomb() over Decomb's own field matching

Decomb has a funny habit of blending everything that it can't properly match, regardless of whether you want it to or not. Or at least, that was the case the last time I checked. It's been a while.

or over using TFM() from the TIVTC plugin?

I literally never heard of that filter until I started this thread so I have no idea.

Well I'd argue that checking the result of an automatic filter also qualifies as doing a diagnostic, but most people would check the source before deciding how it should be handled and which tool to use, and even then they'd check the result.

WTF are you even talking about? What filters count as "automatic" in your mind and which ones don't? What part of "look at the resulting stream frame by frame in Virtualdub" doesn't count as checking the source? What part of "If X, leave the stream alone; if Y, use separatefields" doesn't sound like a post-diagnosis decision-making process?

Throwing your hands up in the air and tossing half the fields away after detecting an occasional orphaned field would probably be most people's definition of not handling a source carefully.

In my experience, there's no such thing as an "occasional" orphaned field.

I was referring to not using "force film".

I know damn well what you were referring to. Your choice of the phrase "unless I'm certain an NTSC DVD is 100% soft telecine", as opposed to "unless an NTSC DVD is 100% soft telecine", strongly implied that you didn't have some kind of in-situ way to check this information for yourself.

Really, you didn't get that I was referring to resolution in respect to picture detail, not picture size? Really, you didn't get that?

I thought you were alluding to the aspect ratio, actually...

There's nothing even remotely "slight" about the judder produced when dropping or repeating frames to convert 23.976 and 29.970 to a common frame rate. It looks terrible.

We'll have to agree to disagree on that, at least when it comes to duplicating frames.

Well not to be picky, but your opening post clearly implies it doesn't matter if the video's not very useful for converting to VCD-compliant MPEG-1, or making anime music videos, or much of anything.

It implied no such thing, and if you think that it does, then either (a) you're hallucinating, or (b) English isn't your first language.

Seriously though..... it's 2016.

Oh wow, it's the current year? Really? I had no idea that it was the current year. It's never been the current year before. This changes EVERYTHING.

http://i.imgur.com/3HHlscK.jpg

Seriously, that's basically a flag for "everything I have to say after this is not worth reading, listening to, or responding to". Thanks for warning me.

I don't have any War Planets DVDs, but I do have the ReBoot Season 3 DVDs released by ADV that OP referred to. Not only is frame repetition obviously the incorrect diagnosis...

No, it's correct. I can tell because avisynth scripts with nothing but mpeg2source and uncomb yielded 100% progressive video streams in which every 5th frame was duplicated.

the suggested Decimate(6) produces the wrong frame rate (24.975 fps).

Well, it's as close as you can get to the correct frame rate without causing A/V desynchronization.

The actual pulldown used by Mainframe Entertainment was direct from 25.000 -> 29.970 without any slowdown. They used a super-long pattern to accomplish this (1200 frames with 199 duplicate fields). The reason I know this is that I manually reverted each episode to 25.000 using a plugin that I wrote, and far too much time.

COOL. What's the name of this plugin and where can it be downloaded? I apparently have some MPEG-1 files to re-encode.

Thinking about it again, that still seems to be standard interlaced NTSC you're describing. It's 59.940 fields per second.

"Standard" NTSC is usually telecined from 24p sources. That's obviously not what I'm talking about.

What you're describing there is still 59.940 fields per second, obtained by first creating full frames, then removing half of each frame later.
Assuming the second method is even likely, I'm not sure how the end result would be any different, or how you'd tell which method was used.

Ummm, you seriously don't understand the difference between 24p and 60i? Okay, I'll give you a tip. Run a video source through Separatefields and look at the result. If you're seeing a repeating pattern of 2 identical frames followed by 3 identical frames (with a bobbing effect), or a 2-2-2-4 pattern, your footage started out as 24p and was later converted to NTSC. If you're seeing 60 easily distinguishable frames per second, then it's natively 60i.

The plugin to which you referred.... is that available for public consumption? :)

LOL, everyone wants ChiDragon's filter now :)

[hello_hello]

Uncomb's lack of cleverness is exactly what makes it such a great diagnostic tool.

Okay.....

If this is achieved by any means other than putting humpty dumpty back together again from his original fields, then it's not an improvement over just scrapping half of the fields.

Wrong.

Decomb has a funny habit of blending everything that it can't properly match, regardless of whether you want it to or not. Or at least, that was the case the last time I checked. It's been a while.

You should check again.

WTF are you even talking about? What filters count as "automatic" in your mind and which ones don't? What part of "look at the resulting stream frame by frame in Virtualdub" doesn't count as checking the source? What part of "If X, leave the stream alone; if Y, use separatefields" doesn't sound like a post-diagnosis decision-making process?

Anything that's done without user intervention is automatic. Even Uncomb(). Does it ask you to confirm every field match?
Yes.... "If X, leave the stream alone; if Y, use separatefields" sounds exactly like a post-diagnosis decision making process.
As does.... "If X, leave the stream alone; if Y, convert the clip to greyscale and flip the picture upside down", and as post-diagnosis decisions go, it's potentially only marginally more silly.


In my experience, there's no such thing as an "occasional" orphaned field.

Okay....

I know damn well what you were referring to. Your choice of the phrase "unless I'm certain an NTSC DVD is 100% soft telecine", as opposed to "unless an NTSC DVD is 100% soft telecine", strongly implied that you didn't have some kind of in-situ way to check this information for yourself.

So you didn't know what I was referring to? Or you did? You knew what I was referring to but you're claiming I implied something else. Are you getting dizzy?

We'll have to agree to disagree on that, at least when it comes to duplicating frames.

We'll have to disagree as to whether we can simply disagree. Duplicating frames to convert 23.976 to 29.970 looks terrible. Dropping frames to convert 29.970 to 23.976 looks terrible too. If you don't agree, you're wrong. If you disagree with that, you're wrong again.

It implied no such thing, and if you think that it does, then either (a) you're hallucinating, or (b) English isn't your first language.

You stated clearly more than once in your opening post, that outputting 60fps progressive was an option. It's there in black and white.
I suggested converting mixed content to 60fps and you poo-poo'd the idea claiming it'd be no good for anything. It's there in black and white.

Oh wow, it's the current year? Really? I had no idea that it was the current year. It's never been the current year before. This changes EVERYTHING.

I haven't seen anyone respond by posting a childish picture in a forum for a quite a while. Isn't it funny how you latched on to a red herring so as not to acknowledge a single point I made when I referred to the year.
Did you expect to be able to mention encoding for VCD and not be reminded how irrelevant that is?

Seriously, that's basically a flag for "everything I have to say after this is not worth reading, listening to, or responding to". Thanks for warning me.

And you think your entire opening post is somehow different??

No, it's correct. I can tell because avisynth scripts with nothing but mpeg2source and uncomb yielded 100% progressive video streams in which every 5th frame was duplicated.

And what do you see without Uncomb? What about just using separatefields? Spot any duplicate frames that way?
What do you think field matching filters do exactly? Do you think they take fields that are already matched up and match them up again? Because that's exactly what you've described. Five frames, repeat one, and for some reason it takes a field matching filter to put the frames back together again.

When pulldown is applied fields are duplicated. That's how it works. The pulldown process results in "dirty" frames. Frames with fields that don't match. The top field from one frame and the bottom from another can end up as a new frame when a single field is duplicated. That's a dirty frame. That's how it works.
When the process is reversed, a field matching filter checks the top and bottom field of a frame, it checks the fields of a frame with the fields of adjacent frames. If it didn't need to do that the fields would already be matched and you wouldn't need field matching in the first place.
So.... when a field is repeated, and the field matching filter tries to match fields from the same frame and also fields from adjacent frames, do you think it's possible at some stage the same field might match twice? And do you think it's possible that'd create a duplicate frame out of thin air? Where do you think the repeated frame comes from when field matching to reverse 3:2 pulldown?
Maybe that's why you think repeating frames to convert the frame rate isn't so bad, because you think there's already repeated frames where there aren't any.

http://avisynth.org.ru/docs/english/externalfilters/uncomb.htm
"UnComb is a simple IVTC filter for matching up even and odd fields of properly telecined NTSC or PAL film source video."
For the sort of non-properly telecined NTSC or PAL you've been complaining about, and from what you've described so far, I'd agree using Uncomb has probably allowed you to correctly diagnose the accuracy of it's help file, but I wouldn't be willing to commit to anything else just yet.

Well, it's as close as you can get to the correct frame rate without causing A/V desynchronization.

No it's not. There's filters that'll decimate frames while aiming for a particular frame rate, as opposed to simply "X" in "Y" decimation.

"Standard" NTSC is usually telecined from 24p sources. That's obviously not what I'm talking about.

No it's not. Standard PAL is 25fps interlaced. Standard NTSC is 29.970fps interlaced.
Telecined 24fps sources are a method of shoe-horning film into an NTSC frame rate by taking advantage of it's 29.970 interlaced nature. Standard NTSC is 29.970 interlaced. That's what it is.

Ummm, you seriously don't understand the difference between 24p and 60i? Okay, I'll give you a tip. Run a video source through Separatefields and look at the result. If you're seeing a repeating pattern of 2 identical frames followed by 3 identical frames (with a bobbing effect), or a 2-2-2-4 pattern, your footage started out as 24p and was later converted to NTSC. If you're seeing 60 easily distinguishable frames per second, then it's natively 60i.

I didn't mention 24p. You seem to have made that connection because you've invented your own definition for NTSC.
NTSC is an interlaced format at 29.970fps. I said that the first time. And the second time.
If I'm seeing 60 unique fields per second it's native 60i. Agreed. That's the definition of NTSC.
You've somehow invented a distinction between standard interlaced NTSC and an imaginary 60fps progressive source where half the fields were removed to make it 60 fields per second instead of 60 frames per second, when it's probably just interlaced NTSC.

[ChiDragon]

The plugin to which you referred.... is that available for public consumption? :)
It is (http://forum.doom9.org/showthread.php?p=1657081&highlight=weaveman#post1657081), but I'm the only person who's ever actually used it.

"Standard" NTSC is usually telecined from 24p sources. That's obviously not what I'm talking about.
The dawn of NTSC television was live-only, with 60 unique pictures per second. 60i TV shows are not and have never been a rarity. It is true that scripted dramas with any sort of budget typically use 24p.

No, it's correct. I can tell because avisynth scripts with nothing but mpeg2source and uncomb yielded 100% progressive video streams in which every 5th frame was duplicated.
Uncomb's entire purpose is to take field duplications and turn them into frame duplications. The correct way to analyze what is truly happening with a source is using only SeparateFields (or Bob) to view the fields as-is. You even mentioned doing this to count field repeats in your post. The distinction between field and frame is the important part here.

[Katie Boundary]

blah blah blah

Awww look at the cute little monkey from PAL-land pretending to understand how NTSC works :)

It is (http://forum.doom9.org/showthread.php?p=1657081&highlight=weaveman#post1657081), but I'm the only person who's ever actually used it.

Expect that number to increase by two.

The dawn of NTSC television was live-only, with 60 unique pictures per second. 60i TV shows are not and have never been a rarity. It is true that scripted dramas with any sort of budget typically use 24p.

Thanks for that info. However, you're underestimating just how much stuff is made in 24p. I might be judder-blind but I was noticing the difference between 24p stuff and 30p stuff back when new episodes of Full House were still being made. "Greg the Bunny", for example, is a half-hour sitcom with no budget beyond what was required for the puppets, but it was done at 24p. Every cartoon ever, up until the use of computer effects, was also done at 24p or even lower frame rates.

[Katie Boundary]

By the way, while you guys have been nitpicking about the ways of dealing with problematic DVDs, I've been adding examples of problematic DVDs, which is what this thread exists for.

[kuchikirukia]

By the way, while you guys have been nitpicking about the ways of dealing with problematic DVDs, I've been adding examples of problematic DVDs, which is what this thread exists for.

You neither know how to inverse telecine nor deinterlace, so your idea of "problematic DVDs" will not really correspond to anyone else's.

But what the heck, I'll join in.
I just picked up a Bluray (DVD upscale) of a series that had one animation cell in one frame in one episode that was combed. The proper way to deal with this series is to pointresize to 480p, separatefields, selecteven or odd (your choice), then bilinearresize to 1080.

[hello_hello]

Awww look at the cute little monkey from PAL-land pretending to understand how NTSC works

Never mind baby Katie. Don't be ashamed you know less about NTSC than a cute little PAL monkey. Stick your fingers in your ears and go back to diagnosing DVDs for us. Share your infinite wisdom. Find us some more of that 60i content created from 60p video by throwing away half the fields. Expose those NTSC DVDs created from PAL by duplicating one frame in six.

By the way, while you guys have been nitpicking about the ways of dealing with problematic DVDs, I've been adding examples of problematic DVDs, which is what this thread exists for.

Way to go baby Katie. Find us some more of that hard telecined content. Put together a list of Hybrid DVDs containing both interlaced NTSC and telecined film. They're so rare I've seen them from PAL-land, but you keep diagnosing girl.

Just hold off handing out ridiculous advice on how to deal with it, otherwise someone could get the impression throwing away half the picture detail is a good idea, or separating fields and leaving them bobbing up and down at 60Hz might be an acceptable way of dealing with DVD content, even if it's not very useful for converting to VCD-compliant MPEG-1, or making anime music videos, or much of anything.

Thanks for that info. However, you're underestimating just how much stuff is made in 24p. I might be judder-blind but I was noticing the difference between 24p stuff and 30p stuff back when new episodes of Full House were still being made. "Greg the Bunny", for example, is a half-hour sitcom with no budget beyond what was required for the puppets, but it was done at 24p. Every cartoon ever, up until the use of computer effects, was also done at 24p or even lower frame rates.

Nobody under-estimated anything baby Katie. Film has been converted to PAL by speeding it up and applying 2:2 pulldown for as long as 3:2 pulldown has existed, but that doesn't mean PAL is sped up film.
NTSC and PAL are both interlaced analogue formats and DVD had to be compatible. Not that I'd take your word for it when it comes to the format of a show, but regardless of the format, if it wasn't NTSC/PAL it had to be shoe-horned in to be compatible with CRT interlaced NTSC/PAL TVs. The majority of content is probably film these days, and much of it ends up on NTSC DVDs using soft pulldown, but "standard" NTSC is 29.970 interlaced frames per second.

[kuchikirukia]

separating fields and leaving them bobbing up and down at 60Hz

separatefields()
a = selecteven()
b = selectodd().turnleft().turnleft().turnleft().turnleft()
overlay(a,b)

[Ghitulescu]

... the ways of dealing with problematic DVDs, ...

There are no problematic DVDs. All of them have been made according to established standards and had to be 100% compliant.

That not all of them are up to the visual quality or that they may not be easily converted into another format (why?), well, that is another problem.

[hello_hello]

There are no problematic DVDs. All of them have been made according to established standards and had to be 100% compliant.
That not all of them are up to the visual quality or that they may not be easily converted into another format (why?), well, that is another problem.

Would you believe "another problem" is what we've been discussing?

[hello_hello]

Has a new category been added or did I miss it the first time?

"Natively NTSC content converted to PAL"
When Film content gets converted to PAL, it typically just gets sped up by 4% and no one notices. However, if something was specifically designed for NTSC and thus contains 30p or 60i content, then the conversion to PAL will absolutely murder it. Expect most if not all of your frames to be blended.
Solution? Buy the region 1 DVDs and use those instead.

QTGMC().SRestore(30) might be worth a try first. DVDs can be expensive.

"Solution? Buy the region 1 DVDs and use those instead. Then you'll have the same nightmares as the rest of us. In fact, this is also what you should do for anything that was originally done at 24 fps".

You're just funin' with us there, aren't you? Buy the DVD again for the privilege of reducing the resolution from 720x576 to 720x480?
Good one....

[Katie Boundary]

You neither know how to inverse telecine nor deinterlace

Actually, I've proven that I can do both. Uncomb/telecide followed by decimate is THE textbook example of proper IVTC, and separatefields is a 100% guaranteed deinterlacer for all but that last category of content.

so your idea of "problematic DVDs" will not really correspond to anyone else's.

The two have nothing to do with each other. You fail logic forever.

But what the heck, I'll join in.
I just picked up a Bluray (DVD upscale) of a series that had one animation cell in one frame in one episode that was combed. The proper way to deal with this series is to pointresize to 480p, separatefields, selecteven or odd (your choice), then bilinearresize to 1080.

That's neither funny, nor remotely similar to how I would deal with the problem, nor appropriate for this thread even if you were being serious (we're discussing DVDs exclusively. Blu-Ray is a different league).

Find us some more of that 60i content created from 60p video by throwing away half the fields.

I'm working on it. We all should be. That's why this thread exists. However, the examples that I provided are proof enough that this phenomenon exists.

Expose those NTSC DVDs created from PAL by duplicating one frame in six.

It's one in five. The decimation ratio afterward is one in six.

Find us some more of that hard telecined content. Put together a list of Hybrid DVDs containing both interlaced NTSC and telecined film.

I assume that by "interlaced NTSC" you mean "Natively NTSC", and yes, I'm on it. I have a rather massive library of DVDs to go through, from "Abyss, The" to "Zoey 101", and it's all going to get categorized.

Just hold off handing out ridiculous advice on how to deal with it, otherwise someone could get the impression throwing away half the picture detail is a good idea

It's better than blending frames.

Also, natively 60i content ALREADY has only half as much picture detail as progressive content, so applying Separatefields doesn't really cost you anything in that situation.

or separating fields and leaving them bobbing up and down at 60Hz might be an acceptable way of dealing with DVD content, even if it's not very useful for converting to VCD-compliant MPEG-1, or making anime music videos, or much of anything.

Hey, I never said that I RECOMMENDED the 60p option. I just said that it might be an option for people who need every ugly blended frame.

Nobody under-estimated anything baby Katie.

A quick look at the amount of 24p content created for broadcast TV proves you wrong on that count.

Film has been converted to PAL by speeding it up and applying 2:2 pulldown for as long as 3:2 pulldown has existed, but that doesn't mean PAL is sped up film.

No sh!t. Have you not read a single thing I've said about War Planets and Reboot?

NTSC and PAL are both interlaced analogue formats and DVD had to be compatible. Not that I'd take your word for it when it comes to the format of a show, but regardless of the format, if it wasn't NTSC/PAL it had to be shoe-horned in to be compatible with CRT interlaced NTSC/PAL TVs. The majority of content is probably film these days, and much of it ends up on NTSC DVDs using soft pulldown

Oooh, please, tell me more things I already learned 12 years ago!

but "standard" NTSC is 29.970 interlaced frames per second.

No. The overwhelming majority of NTSC content was telecined from 24p sources. Native 30p or 60i content is the exception, not the rule, even in shows that include 30p or 60i content.

separatefields()
a = selecteven()
b = selectodd().turnleft().turnleft().turnleft().turnleft()
overlay(a,b)

I don't know what this does but I'm eager to find out.

There are no problematic DVDs. All of them have been made according to established standards and had to be 100% compliant.

You obviously haven't seen the horrors I've seen. Trust me, there's NOTHING standard about some of these DVDs except the fact that they technically play in DVD players.

Has a new category been added or did I miss it the first time?

It has been added.

You're just funin' with us there, aren't you? Buy the DVD again for the privilege of reducing the resolution from 720x576 to 720x480?
Good one....

If it means getting a "less wrong" playback speed and sound pitch, then yes, that's worth a reduction in vertical resolution.

I also added a "buy the region 2 DVDs" option to the "PAL content converted to hard NTSC" category.

[hello_hello]

I'm working on it. We all should be. That's why this thread exists. However, the examples that I provided are proof enough that this phenomenon exists.

No they're not. They're proof that interlaced NTSC exists.

It's one in five. The decimation ratio afterward is one in six.

Sorry but I didn't think describing it as "duplicating one frame in six" would confuse you, or you'd take it to mean duplicating every sixth frame, which would be one in seven.
The duplication ratio is one in six and afterward the decimation ratio is one in six. Unless you're saying should they be different?

Also, natively 60i content ALREADY has only half as much picture detail as progressive content, so applying Separatefields doesn't really cost you anything in that situation.

Really? Temporal resolution isn't a concept with which you're familiar?
Seriously though.... you should have said something sooner. The entire world's been lugging around two fields per frame regardless of the format, oblivious to the fact one field would have sufficed.

Hey, I never said that I RECOMMENDED the 60p option. I just said that it might be an option for people who need every ugly blended frame.

Nope, you also offered it as an option to deal with 60i content. Frame blending not mentioned there.
You never said you RECOMMENDED the 30p option either, which implies both methods are equally satisfactory.
"selecteven() and Selectodd() are optional depending on whether you want 30p or 60p"

A quick look at the amount of 24p content created for broadcast TV proves you wrong on that count.

No it doesn't. The majority of content probably is "film" these days. Nobody has said any different and nobody's under-estimated how much it's used.

Oooh, please, tell me more things I already learned 12 years ago!

So how do you keep those opposing thoughts in your head? You learned what NTSC is 12 years ago, you know PAL isn't sped up film even though that's often the content, but telecine enough film and you think that's standard NTSC.

No. The overwhelming majority of NTSC content was telecined from 24p sources. Native 30p or 60i content is the exception, not the rule, even in shows that include 30p or 60i content.

Really? So for the last 50 years, while everyone's been watching images captured by a TV camera.... while you're watching the news, or watching a game show, or the football, or a soap opera etc.... it was all done with 24p telecining cameras, is that what you're saying?
You can believe standard NTSC is telecined film if it makes you happy, but to the rest of the world NTSC is a 29.970 interlaced format, and telecine is a process by which film is converted to interlaced NTSC.

If it means getting a "less wrong" playback speed and sound pitch, then yes, that's worth a reduction in vertical resolution.

AssumeFPS(24,1) anyone?
Or maybe I should have said AssumeFPS(24000,1001) to convert it back to it's original NTSC format. :)

Not that I've been able to make a lot of comparisons, but in my experience film audio generally isn't pitch corrected when it's sped up for PAL.... maybe because PAL isn't sped up film the way NTSC is telecined film... or something..... but it means when you re-encode the audio, you generally only need to slow it down if you want to go back to the original frame rate. Buying NTSC versions of 24p content is hardly something you should do, or you need to do.

It seems kind of odd in a thread dedicated to dealing with problematic DVDs, I suggested a method of dealing with frame blending a second time and you ignored it a second time.
The topic is still about re-encoding isn't it?

I also added a "buy the region 2 DVDs" option to the "PAL content converted to hard NTSC" category.

No.... apparently not. We've moved on from throwing half the picture away to recommending fixing problematic DVDs by buying better DVDs, although is it possible applying the BuyBluray() plugin would produce even better results?

[Katie Boundary]

Wow, at this point you're literally claiming that natively 60i content doesn't prove the existence of natively 60i content, you're failing basic math (duplicating every sixth frame and then decimating every sixth frame will NOT restore your original frame count), and you're claiming that film isn't 24p.

Go away. Your idiocy is not welcome here.

[hello_hello]

How did you confabulate all that from what I wrote?

You conjured up a theory out of thin air in relation to 60i content. One which I'm sure has no basis in reality because the 60i content you were describing sounds very much like standard NTSC.

Not quite. I'm talking about content where 60 unique frames per second are generated, one field from each frame is discarded, and the surviving fields are then weaved together. I provided some fairly specific examples if you'd like to know what I'm talking about, though you'd have to get the Region 1 versions of the DVDs.

Once again, 60 unique fields weaved together is the definition of NTSC. Of course it exists, whether you want to call it standard NTSC or native NTSC so you can continue believing there's more than one type, but your theory of it originating from 60fps progressive video sounds highly unlikely, and more like one you invented before finally understanding it's probably just plain old interlaced NTSC.

I'll admit it's astounding.... almost admirable....your ability to ignore reality and go with your version of it. I wrote in my previous post:
"I didn't think.... you'd take it to mean duplicating every sixth frame"
yet you replied with:
"duplicating every sixth frame and then decimating every sixth frame will NOT restore your original frame count".
Do you pretend you're replying to what's written and convince yourself it's true later on?

I seriously regret my wording, for no other reason than it gave you another red herring excuse to ignore everything in my post on the actual topic. You have six frames. They were created from five frames. In order to create those six frames one frame in the group of six was duplicated. Not every sixth frame, but if I back-peddle and apologise and rephrase it as duplicating every fifth frame, or call it a duplicated frame in each group of six, can we move on?

....and you're claiming that film isn't 24p

I assume you read my little joke about restoring film back to it's original NTSC frame rate, given your continued insistence standard NTSC is telecined film, and despite the fact I described restoring it back to 24p first, and despite the smiley that followed my little joke, you're red-herring it into a new topic avoiding tangent. I can't imagine anyone with any amount of common sense taking that literally, but it seems you did, or you're pretending to.

[Katie Boundary]

Allow me to repeat:

Go away. Your idiocy is not welcome here.

EDIT: There is one condition under which I will continue to indulge your nonsense. You must buy or rent the Region 1 version of Where on Earth is Carmen Sandiego? on DVD, rip the first chapter of the first episode, apply Separatefields, go through the C5 Corridor scene in the opening credits frame-by-frame in Virtualdub, and then apologize and admit that you were wrong and I was right about this type of content existing. Then, and only then, will I consider your posts even worth reading, much less responding to.

[wonkey_monkey]

Go away. Your idiocy is not welcome here.

You are not a mod. They decide who is welcome, and they decide which posts get stickied. Apparently yours hasn't meant the criteria yet.

Going back to your opening post:

however, if there are still a few interlaced frames because an incompetent telecine job left orphaned fields, you'll need to scrap Uncomb and replace it with Separatefields(), followed by either Selecteven() or Selectodd(), followed by decimate().

That's a stupid thing to do and if you can't understand why, you shouldn't be pontificating in this fashion.

60i content

While almost no shows have been done entirely in 60i since the end of the Cold War, many of them at least include short special effects that are rendered in this way. Fades, especially in opening credits or title sequences, are also frequently done in 60i. The ONLY way to clean up this footage is to use Separatefields(); selecteven() and Selectodd() are optional depending on whether you want 30p or 60p.

Nonsense. separatefields is not the right tool for the job. Please stop saying it is. At the very least you should understand the difference between separatefields() and bob().

Captured content from VHS or Beta

The second-ugliest of all kinds of footage. Using separatefields() reveals that this is not just effectively 60i, but that each pseudo-field is made up of two of the original film frames, blended unevenly from top to bottom.

I don't even know what you're talking about. "Each pseudo-field is made up of two of the original film frames"? That's nonsense. It may be the case with some transfers, but it's not an intrinsic property of analogue video, or of digital captures of analogue video.

If you've got 60i on analogue video, you can capture it at 60i. If you've got a movie on NTSC analogue video, then it will probably just be hard telecined. I don't know where you've got this rubbish about blended frames from. You probably screwed up the capture.

[Katie Boundary]

At the very least you should understand the difference between separatefields() and bob().

(looks at the Bob documentation)

Holy crap! Why did I never notice this before? Thank you. The OP will be edited accordingly.

I don't even know what you're talking about. "Each pseudo-field is made up of two of the original film frames"? That's nonsense.

"Nonsense" as in "what the hell was Shout Factory thinking when they did this"? Absolutely. "Nonsense" as in "this has never happened"? Haha, I wish. I've seen this content. It exists. I gave examples.

https://scontent-dfw1-1.xx.fbcdn.net/hphotos-xtp1/v/t1.0-9/13010876_1679495415648247_3078440992632956785_n.jpg?oh=d2ce8e1da755dc12036e47ce3fdb07f9&oe=57AA3DB1

See that? I didn't do that. The idiots at Cookie Jar and/or Shout Factory did that. And this isn't even the worst capture job I've seen. Don't make me dig up my old Evangelion DVDs.

It may be the case with some transfers, but it's not an intrinsic property of analogue video, or of digital captures of analogue video.

Agreed.

If you've got 60i on analogue video, you can capture it at 60i. If you've got a movie on NTSC analogue video, then it will probably just be hard telecined. I don't know where you've got this rubbish about blended frames from. You probably screwed up the capture.

Dude, I don't even have a f***ing capture card! This is straight from the DVD!

[wonkey_monkey]

(looks at the Bob documentation)

Holy crap! Why did I never notice this before? Thank you. The OP will be edited accordingly.

Bob is one of the simplest, earliest filters anyone going anywhere near interlaced content should comprehend.

The fact that you've never even heard of it should, at least, give you pause to reconsider your self-appointed title of Interlacing Expert.

I've seen this content. It exists. I gave examples.

The problem is that you implied that this was the case with ALL captures from VHS/Beta.
______________________________________

From your edited first post:

you'll need to scrap Uncomb and replace it with Bob(), followed by either Selecteven() or Selectodd(), followed by decimate(). If you're downscaling to 288p or less, you can use Separatefields instead of Bob() to save a few CPU cycles.

*facepalm* Please, please stop trying to give out advice. You don't know what you're doing. Just because you've now learnt about bob(), doesn't mean you have all the answers.

Firstly, there are myriad ways to solve this problem; earlier, you understood one very bad way. Now, you almost understand another way, but it's almost as bad as the first way. There are a number of much better ways.

Secondly, no, you can't simply use separatefields() instead of bob() if you're downscaling.

You do not understand what you're trying to teach. Please stop teaching.

[Katie Boundary]

Bob is one of the simplest, earliest filters anyone going anywhere near interlaced content should comprehend.

The fact that you've never even heard of it should, at least, give you pause to reconsider your self-appointed title of Interlacing Expert.

My expertise is in the different ways in which a DVD transfer can get completely borked, and how to diagnose these transfers. I never claimed to be an expert at repair.

The problem is that you implied that this was the case with ALL captures from VHS/Beta.

No, I didn't. An uncomb diagnosis won't even detect properly captured content.

Please, please stop trying to give out advice. You don't know what you're doing.

I absolutely do. And this isn't even an advice thread. It's a warning thread. My suggestions for how to deal with problem content are mere sidenotes to the real focus, which is a list of DVDs with screwed-up transfers along with detailed explanations of the ways in which they are screwed up.

Just because you've now learnt about bob(), doesn't mean you have all the answers.

I never claimed to.

Firstly, there are myriad ways to solve this problem; earlier, you understood one very bad way. Now, you almost understand another way, but it's almost as bad as the first way. There are a number of much better ways.

I don't doubt that there are other ways. What I doubt is that they're much better.

Secondly, no, you can't simply use separatefields() instead of bob() if you're downscaling.

Umm... you totally can. You won't get the same results, but some people might accept the slight increase in aliasing to save a few CPU cycles.

[wonkey_monkey]

My expertise is in the different ways in which a DVD transfer can get completely borked, and how to diagnose these transfers. I never claimed to be an expert at repair.

You imply your "expertise" through the use of phrases like "The best way to deal with this kind of content is..." and "you'll need to..." and "The best solution is to..." and "The ONLY way to clean up this footage is to..." and "your only safe bet is..."

None of which are accurate, by the way.

The problem is that you implied that this was the case with ALL captures from VHS/Beta.
No, I didn't.

Yes, you absolutely did. I see you've now edited your post to talk about "Badly captured content" but you are still actually only talking one particular kind of bad "capture". In fact, this problem is not in any way limited to, exclusive to, or even particularly related to VHS/Beta, or any other kind of "capture" at all.

It's just a different example of "bad" framerate conversion and could be performed on any kind of video.

Please, please stop trying to give out advice. You don't know what you're doing.
I absolutely do.

You demonstrably do not. You hadn't even heard of bob() before I told you an hour ago, and your advice is still wrong even after you've edited it. You do not understand the consequences of using separatefields() and when it's appropriate to use it.

And this isn't even an advice thread.

I refer you, again, to these quotes:

The best way to deal with this kind of content is...
you'll need to...
The best solution is to...
The ONLY way to clean up this footage is to...
your only safe bet is...

...a list of DVDs with screwed-up transfers along with detailed explanations of the ways in which they are screwed up.

No-one needs such a list, and your explanations are not detailed at all.

Just because you've now learnt about bob(), doesn't mean you have all the answers.
I never claimed to.

Do I need to reiterate that list of quotes a third time? Your very words betray you as you literally claim to have the answers.

I don't doubt that there are other ways. What I doubt is that they're much better.

They are much, much, better. Take it from someone who worked in the engineering department of a TV station for ten years, and used to be given all the NTSC->PAL and PAL->NTSC conversion jobs because I could do them better than the slightly aged Avid edit suites?

Umm... you totally can. You won't get the same results, but some people might accept the slight increase in aliasing to save a few CPU cycles.

No, you can't, and it has nothing to do with aliasing. The fact that you don't even know what the problem is demonstrates, once again, just how uninformed you are.

[Katie Boundary]

You imply your "expertise" through the use of phrases like "The best way to deal with this kind of content is..." and "you'll need to..." and "The best solution is to..." and "The ONLY way to clean up this footage is to..." and "your only safe bet is..."

The fact that you hallucinated subtext into my questionable choice of words is not evidence that I was implying anything. Did you think that I was also literally suggesting making a deal with Satan for that last category of content?

Yes, you absolutely did. I see you've now edited your post to talk about "Badly captured content" but you are still actually only talking one particular kind of bad "capture".

This would be an excellent time for you to provide examples of DVDs with other kinds of screwed-up captures.

In fact, this problem is not in any way limited to, exclusive to, or even particularly related to VHS/Beta, or any other kind of "capture" at all.

When the blending is uneven from top to bottom, as is the case with the Evangelion DVDs, that's the kind of error that can only result from using a capture card. Screwing it up by any other method would require either a lot more work and a deliberate intent to screw it up that way, or would result in even worse problems than what I described.

It's just a different example of "bad" framerate conversion and could be performed on any kind of video.

In the case of Captain Planet, where the frame-blending is even from top to bottom, you're right. I will consider renaming that category of problem.

You demonstrably do not. You hadn't even heard of bob() before I told you an hour ago

Oh, I had heard of it. I just didn't read the documentation to figure out what it did, or more likely, I read the documentation many years ago but forgot about it in favor of separatefields because separatefields was mentioned in the documentation for other filters and bob wasn't. Either way, it has no bearing on the fact that from the beginning, I was aware of the disadvantages of every restoration method that I proposed.

and your advice is still wrong even after you've edited it. You do not understand the consequences of using separatefields()

YES, I DO. There is a doubling of frame rate and frame count, a halving of vertical resolution, a slight aliasing effect identical to what you'd get from a pointresize, and a 60hz bobbing effect from the fact that the top and bottom frames are no longer occupying slightly different heights. None of this is news to me. I have ALWAYS regarded Separatefields as the "nuke the site from orbit" option for those reasons.

No-one needs such a list

Demonstrably wrong. I know that such people exist because I was one of them many years ago.

and your explanations are not detailed at all.

They're as detailed as they need to be.

Do I need to reiterate that list of quotes a third time? Your very words betray you as you literally claim to have the answers.

A questionable choice of words is not the same as a claim about my own qualifications. The only times when I mention myself at all are when I say things like "I first noticed this problem when..."

They are much, much, better.

Feel free to elaborate.

Take it from someone who worked in the engineering department of a TV station for ten years, and used to be given all the NTSC->PAL...

You have my sincerest sympathies. I don't think that I could ever bring myself to commit the atrocities necessary to convert NTSC to PAL, unless the NTSC content was originally telecined from film and I could just IVTC it back and then do a film-to-PAL job.

No, you can't, and it has nothing to do with aliasing.

Yes, you can. Watch:

mpeg2source("101.d2v")

separatefields()

bilinearresize(352,240)

Then load that into Virtualdub and notice as it gives you images instead of an error message! Whether you like the results or not is a different story.

[Stereodude]

http://i.imgur.com/q9uxuOF.jpg

[manolito]

Amen...

Can a mod please close this thread?

[hello_hello]

This is comedy gold. Well it would be if it wasn't kind of sad. It doesn't matter how may people tell Katie she's wrong and how many different ways they try to explain it, Katie still knows best. If nothing else at least the thread's a little entertaining.

EDIT: There is one condition under which I will continue to indulge your nonsense.

Wrong again. It's pretty simple. When you make a claim in a forum, especially one that goes against the norm, as is the case in the real world, the onus is on you to prove what you're saying is correct. Upload a sample.

I still have absolutely no idea how you're looking at interlaced video and coming to the conclusion it must have begun it's life as progressive video at twice the frame rate, with half the fields removed for interlaced NTSC or 60i. How do you know? Please explain.

The fact that you hallucinated subtext into my questionable choice of words is not evidence that I was implying anything. Did you think that I was also literally suggesting making a deal with Satan for that last category of content?

That's funny coming from someone who only a few posts ago accused me of claiming film wasn't 24p. Katie's the queen of hallucinating subtext when she needs to create a tangent to wander off on.

You'd imagine nobody would offer advice in the form of "the best way" or "the only way", or "your only safe bet", argue about it relentlessly when questioned, then later expect to get away with a packpeddle as feeble as "I never claimed to be an expert".

I have ALWAYS regarded Separatefields as the "nuke the site from orbit" option for those reasons.

I laughed and I laughed and I laughed and I laughed......

Step1: Uncomb()
Step2: Nuke.

The way history is slowly being re-witten with each edit of the opening post is amusing. I wish I'd taken a snapshot of the alpha version. I remember suggesting treating mixed frame rate NTSC as variable frame rate and the idea was immediately poo-poo'd. Now it's "even preferable for certain extremely specific purposes". I don't know what the definition of an extremely scientific video conversion might be. Getting it right, I suspect.

It seems like only yesterday the self proclaimed expert at diagnosing the different ways DVD transfers can get completely borked thought "PAL Hardcoded as NTSC" DVDs were created the following way and argued about it in several posts.

Step 1: find PAL/SECAM content.
Step 2: duplicate every 5th frame instead of every 4th.
Step 3: Interlace.

Oh that's right.... it was only yesterday, or maybe the day before. And while the penny hasn't completely dropped and the opening post still refers to duplicated frames, not fields, apparently I'm the one posting nonsense our resident self proclaimed expert at diagnosing DVD transfers is forced to tolerate. Pfffffttt.....

Any chance history will be re-written a little again to allow re-encoding PAL film transfers with AssumeFPS(24,1) or will replacing PAL film transfers with NTSC DVDs remain the Katie recommended option for re-encoding?

[Stereodude]

If you keep feeding it...

[Katie Boundary]

hello is now on my ignore list. I apologize for allowing him/her to derail this thread so badly and for so long.

Now, can we PLEASE get back to the actual subject matter, i.e., compiling a list of DVDs that do not respond well to the standard "match fields and decimate by five" approach to IVTC?

[hello_hello]

If you keep feeding it...

Yeah I know..... but apparently Katie's enabled the forum "can't be wrong" function so my posts can't effect her level of detachment from reality.

hello is now on my ignore list. I apologize for allowing him/her to derail this thread so badly and for so long.

It seems sticking her fingers in her ears means Katie can't be wrong, so I imagine reflecting her own behaviour on someone else and apologising for it on their behalf is just another symptom of the same underlying condition.

When someone starts a thread and nearly a week later every single poster has disagreed with them.... how does a rational adult not consider the possibility it's not the rest of the world getting it wrong?

Now, can we PLEASE get back to the actual subject matter, i.e., compiling a list of DVDs that do not respond well to the standard "match fields and decimate by five" approach to IVTC?

I wonder when Katie will realise if it's okay for her to offer advice on how to handle DVDs, it's okay for others to do the same?
I wonder if she'll ever understand her definition of the "standard match fields and decimate by five approach" is just one she invented before she understood how field matching filters work, and it's unlikely anybody else here uses her method, given Uncomb() is a prerequisite for membership of that particular club?
Unfortunately though, it seems unlikely Katie and reality will be meeting again any time soon....

[wonkey_monkey]

YES, I DO. There is a doubling of frame rate and frame count, a halving of vertical resolution, a slight aliasing effect identical to what you'd get from a pointresize, and a 60hz bobbing effect from the fact that the top and bottom frames are no longer occupying slightly different heights. None of this is news to me. I have ALWAYS regarded Separatefields as the "nuke the site from orbit" option for those reasons.

Then why do you keep recommending it when it's the worst possible way to fix anything?

They are much, much, better.
Feel free to elaborate.

I don't see why I should when you've already told everyone what's best. I mean, literally. You have been literally telling people that "the best way to fix this is..." and then giving them a terrible suggestion. You've even changed your "best way" and "only way" suggestions so you clearly didn't know what was "best" before, and you still don't.

Then load that into Virtualdub and notice as it gives you images instead of an error message!

Oh for... I meant that you can't do it and expect a correct result. You might as well just recommend that people do nothing and put up with the judder and interlaced frames.

Literally everyone other person who has commented on this thread has done nothing but tell you how wrong you are, and yet you still refuse to even consider the idea.

[Ghitulescu]

I thought Katie would have been since long recognized as a troll and subsequently thrown away the very next second.

I was wrong.

[manolito]

Please someone tell me that this thread isn't really about DVD transfers, but in reality it is part of a gender study. Like: How do gay women prevail in a field which is dominated by males? Some of Katie's posts sound like they have been taken from a psychology textbook...

[hello_hello]

PAL/SECAM content encoded as hard NTSC
I first learned about this phenomenon when playing around with ripped War Planets DVDs. Solution: Uncomb() followed by Decimate(6). Or just get the Region 2 discs like a civilized person. In the extremely specific case of shows made by Mainframe Entertainment, which actually duplicated 199 out of every 1000 frames (or whatever) instead of speeding up by 0.1% and then duplicating 1 out of every 5, you can use Chidragon's plugin to avoid accidentally deleting every thousandth frame.

Please someone correct me if I'm wrong, but when you speed something up it gets faster, yes?

Better check....
25fps x 101% = 25.25 x 6 / 5 = 30.30fps
No, that's too fast, so I guess it's slowed down, not sped up.

I know it's an easy mistake to make and under most circumstances fairly forgivable, but not for a self proclaimed expert on diagnosing the different ways a DVD transfer can get completely borked.

[Stereodude]

Please someone tell me that this thread isn't really about DVD transfers, but in reality it is part of a gender study. Like: How do gay women prevail in a field which is dominated by males? Some of Katie's posts sound like they have been taken from a psychology textbook...
Someone is intentionally trolling us and having a field day with it as some of the posters here can't resist trying to convince "her" that she's wrong.

[Groucho2004]

Someone is intentionally trolling us and having a field day with it as some of the posters here can't resist trying to convince "her" that she's wrong.
More or less my thoughts as well. Don't know if it's trolling or boredom, he/she sure pushes some people's buttons. :D

Please someone tell me that this thread isn't really about DVD transfers, but in reality it is part of a gender study. Like: How do gay women prevail in a field which is dominated by males? Some of Katie's posts sound like they have been taken from a psychology textbook...
Don't let monikers fool you.

Edit - Or maybe this (http://www.amazon.co.uk/gp/pdp/profile/A2JPFU60X8KYE2/ref=cm_cr_rdp_pdp) is her?

[Katie Boundary]

Well... last night, I was dissecting the first two episodes of The Secret world of Alex Mack, and found some pretty interesting stuff. The live-action footage was done at 24 fps, as expected. The animated titles, simulated screen images, and computer-generated "puddle" effects were 30p. A particular green-screen shot in episode 2 was done at 60i. And then, maybe 2/3 of the way into the pilot episode, I noticed something particularly trippy: 48i footage! I'm not sure if it was originally shot at 48 fps, or if it was shot at 24 and then cranked up to 48, but either way, every fourth frame was duplicated afterward to yield 60p, and then each resultant frame contributed only one field to the final stream. I was left wondering "why would anyone do that?"

Then why do you keep recommending it when it's the worst possible way to fix anything?

That's like asking me why I drink tabasco sauce when Hitler is a unicorn. Absolutely none of it is true.

I don't see why I should when you've already told everyone what's best. I mean, literally. You have been literally telling people that "the best way to fix this is..." and then giving them a terrible suggestion. You've even changed your "best way" and "only way" suggestions so you clearly didn't know what was "best" before, and you still don't.

Oh for f*ck's sake, slight overstatements on my part do NOT give you carte blanche to make personal attacks without evidence!

Oh for... I meant that you can't do it and expect a correct result.

OH. Well, that's a different thing entirely :)

Literally everyone other person who has commented on this thread has done nothing but tell you how wrong you are, and yet you still refuse to even consider the idea.

Except for that one time when I was actually proven wrong about something, admitted it, and fixed the OP accordingly.

Please someone tell me that this thread isn't really about DVD transfers, but in reality it is part of a gender study. Like: How do gay women prevail in a field which is dominated by males?

1) gay autistic women
2) gender studies classes and the entire field of psychology are bullcrap
3) Do you have anything relevant to contribute? If not, STFU and GTFO.


Edit - Or maybe this (http://www.amazon.co.uk/gp/pdp/profile/A2JPFU60X8KYE2/ref=cm_cr_rdp_pdp) is her?

Damn... you found me.

[hello_hello]

Someone is intentionally trolling us and having a field day with it as some of the posters here can't resist trying to convince "her" that she's wrong.

Maybe.... but I'm not sure. I think there's a reasonable chance she's genuinely not playing with a full deck.
Either way though, I'll confess her entertainment value has diminished considerably. Actually, she's now pretty much just plain old boring. http://forum.doom9.org/showthread.php?p=1765177#post1765177

[wonkey_monkey]

Then why do you keep recommending it when it's the worst possible way to fix anything?
That's like asking me why I drink tabasco sauce when Hitler is a unicorn. Absolutely none of it is true.

But you do keep recommending it - originally five, now four times in the OP. And it is the worst possible way to fix the things you're trying to fix. So that's one true thing, and another true thing. I make that... two true things.

Oh for f*ck's sake, slight overstatements on my part do NOT give you carte blanche to make personal attacks without evidence!

I have not made any personal attack. I have attacked your statements. I have not, for example, condescendingly referred to anyone as a "cute little monkey from PAL-land."

Declaring a bad solution to the "best way" is not a "slight overstatement." It's wrong.

3) Do you have anything relevant to contribute? If not, STFU and GTFO.

I do wish you'd take your own advice. Have you ever considered that reason everyone disagrees with you is not because you're right and they're wrong?

Literally everyone other person who has commented on this thread has done nothing but tell you how wrong you are, and yet you still refuse to even consider the idea.
Except for that one time when I was actually proven wrong about something, admitted it, and fixed the OP accordingly.

If you're talking about replacing separatefields() with bob(), then no, you did not "fix" the OP. You've just gone from recommend the very worst way of fixing something to recommending the second-worst way.

[Katie Boundary]

Note: this is the last time when I will respond to a post that is not supported by any examples or evidence. From now on, the rule is "put up or shut up".

But you do keep recommending it - originally five, now four times in the OP.

If you believe that, then your definition of "recommend" is different from the one that the rest of the English-speaking world uses.

And it is the worst possible way to fix the things you're trying to fix.

Wrong. Blending is worse.

I have not made any personal attack. I have attacked your statements.

"You don't know what you're doing" is a personal attack, not an attack on my statements.

I have not, for example, condescendingly referred to anyone as a "cute little monkey from PAL-land."

Personal attacks are fine as long as they're supported by evidence :) In this case, I had provided more than enough evidence to prove hello wrong on pretty much every count of what he/she was saying.

Declaring a bad solution to the "best way" is not a "slight overstatement." It's wrong.

I see your OCPD is as big as mine.

https://s3.amazonaws.com/assets.forward.com/images/cropped/schwartz-awakens-1425564293.jpg

Some problems are so bad that they only have bad solutions. The difference between "best solution" and "least bad solution" is a purely linguistic one at that point. 60i content is such a problem.

While the bob filter's "shimmering" is normally unnoticeable or barely noticeable, it tends to be quite problematic during shots of outer space because stars often occupy only one or two pixels and the filter will cause them to flicker at 30 hz (a problem that I first noticed with separatefields). I'd be extremely interested in a bob-deinterlacer that did both vertical and temporal interpolation to solve this problem, even if it resulted in slight combing at scene changes.

Have you ever considered that reason everyone disagrees with you is not because you're right and they're wrong?

I always consider the possibility. And with one exception, I've rejected it on the basis that either the opposing claims contradicted all available evidence, were not supported by the available evidence, or just didn't make any kind of semantic sense.

If you're talking about replacing separatefields() with bob(), then no, you did not "fix" the OP. You've just gone from recommend the very worst way of fixing something to recommending the second-worst way.

I went from the second-least bad to the least bad.

Of course, if there's an objectively better filter for this purpose than bob(), then you could have shut down this entire line of discussion days ago with a single link.

EDIT: never mind that comment about temporal interpolation. It would just be poorly disguised frame-blending. I'm not sure why I didn't realize that earlier.

[hello_hello]

Could somebody please inform the forum expert on diagnosing the different ways a DVD transfer can get completely borked... she wanted samples so there's some in a zip file attached to this post if she cares to take her blinkers off to have a look.

Note: this is the last time when I will respond to a post that is not supported by any examples or evidence. From now on, the rule is "put up or shut up".

Aside from the obvious delusion of grandeur, you're once again proving your hypocrisy. Where's the samples I've asked for?

So there's samples attached.... although someone might have to point it out, given you've enabled the forum "determined to remain ignorant" function so you can't be wrong. If they do, and "she who can't be told" looks at the samples, "she who won't be wrong" will be expected to practice what she pontificates. Lets start with a sample of video for which IVTC can't be applied and the only option, according to "the self proclaimed expert on diagnosing the different ways a DVD transfer can get completely borked", is Bob() followed by Selecteven().Decimate()

It'll be Katie's turn to put up or shut the hell up and spare us the pain.

While the bob filter's "shimmering" is normally unnoticeable or barely noticeable, it tends to be quite problematic during shots of outer space because stars often occupy only one or two pixels and the filter will cause them to flicker at 30 hz (a problem that I first noticed with separatefields). I'd be extremely interested in a bob-deinterlacer that did both vertical and temporal interpolation to solve this problem, even if it resulted in slight combing at scene changes.

Imagine if you hadn't spent the last five days pontificating and listened instead?

Of course, if there's an objectively better filter for this purpose than bob(), then you could have shut down this entire line of discussion days ago with a single link.

It took days of arguing to get to you move past separatefields() and onto something resembling de-interlacing, and days of arguing before you came to understand what interlaced NTSC is, assuming you finally got there. You only managed to get as far as Bob() yesterday, and now you're claiming to have an open mind??? Comedy gold!!

If only I'd mentioned there's better methods in the first place. Oh.... wait a minute...... remember the furious arguing this started a week ago?
"60i content"? Isn't that the definition of NTSC interlaced video? There's de-interlacers for that sort of thing, rather than throwing away half the fields.

The samples are PAL. Hopefully even "she who knows everything" will realise exactly the same principle applies to de-interlacing PAL as it does for NTSC, so we won't need to spend days arguing about that one before she understands.
Small steps though, let's see if Katie can at least accept there's other ways of doing things, even if she doesn't find them all that much better. Let's hope she finally understands her claim she's suggesting the ONLY way to clean up 60i footage is obviously wrong.

Katie knows best.zip (http://www.filedropper.com/katieknowsbest) (13MB)