VideoHelp Forum
+ Reply to Thread
Results 1 to 15 of 15
Thread
  1. Member
    Join Date
    Mar 2017
    Location
    Salem, MA
    Search Comp PM
    Hi! New member, first post. I'm looking forward to picking your brains about something odd.

    Project background: my source is an interlaced NTSC DVD. It was shot on film (24fps) but edited on video (29.97), so it's interlaced and has broken cadences throughout. I can't use an automated IVTC process to recover the original frames.

    My goal is to extract short 30-second clips, combine the interlaced fields back to the original full frames, and upscale to 720p or 1080p. Because of the eventual upscale, I'm trying to restore the frames losslessly -- or as close to lossless as I can realistically manage.

    Because my clips will be short, I've been looking at various ways to deinterlace *manually*. Here's where it gets weird...


    Originally, I used Streamclip to deinterlace the DVD's VOB. Outputting at 23.976fps left me with some skipped/duplicated frames thanks to the VOB's broken cadences, so I began outputting at 29.97fps instead. This gave me a result that had NO skipped frames -- but a repeated frame every five frames. Its position would change every time there was a cadence break, but I'd just go through manually and delete the dupes. It seemed to work well for my purposes.

    But I just learned that when Streamclip deinterlaces -- it throws away half the fields and interpolates the missing data! I was losing half my image data and a lot of sharpness -- and that's no good for an eventual upscale. So now I'm looking into ways to make Streamclip KEEP both fields.


    It looks like if I set Streamclip to "DO NOT deinterlace", and set the output to 29.97fps, the result maintains the VOB's current interlacing -- just combining every two adjacent fields into a frame. I get three clear frames, then two interlaced (toothy) frames. Obviously the two toothy frames are a problem -- but their existence seems to indicate that Streamclip is now KEEPING and COMBINING both fields, not throwing one set away.

    So, first question: do I have that right? With these settings, is Streamclip KEEPING both sets of fields? Is the image data now lossless (if occasionally interlaced)?


    If I'm right -- here's where my question gets weirder. If Streamclip CAN recombine two fields to losslessly recreate a film frame -- can I force it to recombine the fields DIFFERENTLY, to recreate the frame split across the two toothy frames?

    A little visual aid: let's take a four-frame film clip and follow it through all the steps I'm talking about.

    It starts life as four solid film frames, at 24fps:
    1, 2, 3, 4

    It's converted to interlaced video at 29.97fps and put on the DVD. It's now made up of ten fields that, when played back, run in this order. (A is upper field, B is lower field):
    1a, 1b, 1a, 2b, 2a, 3b, 3a, 3b, 4a, 4b

    When Streamclip deinterlaces, and outputs at 29.97 instead of 23.976, it gives me full frames that are interpolated from each of the upper fields:
    1a, 1a, 2a, 3a, 4a

    I don't lose any frames, and I get a single duplicate -- but I lose image data on each frame. Boooo.

    When Streamclip does NOT deinterlace, and outputs at 29.97, it looks like it combines every two adjacent fields to give me these five frames:
    1a1b, 1a2b, 2a3b, 3a3b, 4a4b

    The first, fourth and fifth video frame have correctly recombined the fields to restore film frames 1, 3 and 4. But film frame 2 is still split across the toothy video frames 1a2b and 2a3b.


    So I had the thought: can I make Streamclip recombine the OTHER adjacent fields? Instead of having it combine the first-second, third-fourth, fifth-sixth fields on the VOB -- can I have it combine the second-third, fourth-fifth, sixth-seventh, and so on?

    If I could do that, then I assume Streamclip would take those same ten fields and generate six frames, like this:
    blank-1a, 1b1a, 2b2a, 3b3a, 3b4a, 4b-blank

    This process would restore film frames 1-3, but frame 4 would be split across video frames 3b4a and 4b-blank. (But that's okay, because the previous process restored frames 1, 3 and 4!)


    If I could make Streamclip do this, then by using both methods on the same clip I could restore ALL the film frames -- and then I could throw out the duplicates and the toothy frames, and be left with the original film frames 1-4 -- with no deleted fields, no interpolation and no data loss.


    So, the big question: CAN Streamclip be told to combine the other adjacent fields? I tried changing its setting from "upper field first" to "lower field first", but that did nothing. It continued to recombine the exact same sets of fields as when I had "upper field first" selected.

    If not -- is there another Mac-based program that can?

    (I've been told that AviSynth can separate and combine fields, but it's PC-based and script-based, and both of those things confuse me to death...)


    I realize that this is a CRAZILY time-intensive way of deinterlacing -- but I'm trying to keep from losing ANY image data or deleting any fields, and I don't know of any other way that doesn't involve interpolation. And because my clips are short I don't mind putting in this time.

    I look forward to hearing what you gurus have to say! Thanks!
    Last edited by jephyork; 22nd Mar 2017 at 14:42.
    Quote Quote  
  2. Originally Posted by jephyork View Post


    So, first question: do I have that right? With these settings, is Streamclip KEEPING both sets of fields? Is the image data now lossless (if occasionally interlaced)?
    Yes .



    If I'm right -- here's where my question gets weirder. If Streamclip CAN recombine two fields to losslessly recreate a film frame -- can I force it to recombine the fields DIFFERENTLY, to recreate the frame split across the two toothy frames?
    No mpegstreamclip cannot

    What you want is adaptive IVTC (inverse telecine) , ie. to field match and decimate adaptively over a range of fields , giving you back the original film frames in full resolution

    If it was a perfect 3:2 pattern, you're done. But you actually want something more advanced if there are cadence breaks, because there could be orphaned fields. If a remaining single field from a film frame has no matching field, it will have combing, so in those scenarios that individual field is interpolated . That is where avisynth gives you more, and higher quality options instead of a "dumb" field resize which will leave that frame with aliasing and jaggy edges

    So, the big question: CAN Streamclip be told to combine the other adjacent fields? I tried changing its setting from "upper field first" to "lower field first", but that did nothing. It continued to recombine the exact same sets of fields as when I had "upper field first" selected.
    Nope;

    If not -- is there another Mac-based program that can?

    (I've been told that AviSynth can separate and combine fields, but it's PC-based and script-based, and both of those things confuse me to death...)
    If you can't access a PC which really is the "best" way (go to a friends house for example) , vapoursynth is available on OSX, but there is a bit of a learning curve, even more than avisynth

    Other Mac options are :

    1) mac version of avidemux using "decomb telecide" then "remove IVTC dupe" in that order. avidemux is GUI based which is a plus for most people, but warning it can be buggy and prone to crashing
    2) ffmpeg compiled for OSX using -vf fieldmatch=order=tff:combmatch=full,yadif=deint=in terlaced,decimate (or bff if bottom field first)

    BUT - Even though both were ported from their avisynth counterparts, for some reason they give worse results. Sometimes they miss something and there is a "jump" , when the exact same filters in avisynth or vapoursynth get it correct. And that' s not a "Mac" issue, because linux and windows versions of avidemux and ffmpeg also have the same issues
    Quote Quote  
  3. Member
    Join Date
    Mar 2017
    Location
    Salem, MA
    Search Comp PM
    Thanks for the response! This is helpful.

    Okay -- so I can make Streamclip give me 3/4 of the film frames I need, with no quality loss -- but it has no ability to flip the fields and get me that last frame. That's good to know!

    (Thanks for reminding me of orphaned fields! I figured that I'd encounter a few of those at starts/ends of shots, and was planning to just interpolate them to full frames using whatever the highest-quality software for that is.)

    It sounds like I'm going to have to learn a little AviSynth after all. It's not so much that I don't have access to a PC -- it's just that I don't really know how to run them! I've been told that I can install Wine on my Mac and run AviSynth that way ... I was hoping to avoid jumping into a pool that I don't know how to swim in, but I guess I'll have to learn!

    Still, I'd like to minimize my time in PC-land, and do most of my work on systems and software that I'm familiar with. Here's my idea for how to proceed:

    - Use Streamclip to generate the clip I need, still interlaced at 29.97, so I have all the fields. (As I described above.)
    - Get AviSynth up and running, and toss that clip into it.
    - Separate all the fields.
    - Delete the very first field, and the very last field, of the clip. (I'll add a junk frame at the start and end so I don't lose anything I need.)
    - Weave the fields back together.

    This should get me what I need. Right?

    Using my 4-frame example above, I'd have Streamclip generate video frames consisting of:
    1a1b, 1a2b, 2a3b, 3a3b, 4a4b

    Then have Avisynth separate them into:
    1a, 1b, 1a, 2b, 2a, 3b, 3a, 3b, 4a, 4b

    Then have AviSynth delete the first and last field, giving me:
    1b, 1a, 2b, 2a, 3b, 3a, 3b, 4a

    Then have AviSynth weave what's left back together into:
    1b1a, 2b2a, 3b3a, 3b4a

    Giving me back all film frames except 4. (Which I'd already have in my initial Streamclip file.)


    Is my thinking right? That AviSynth (a) can delete individual fields, and (b) would recombine the results this way?


    Also: can it do this losslessly? And what format/codec would the original Streamclip file need to be, to work in AviSynth? (I usually transcode the VOBs into ProRes44 mpegs, which I think is lossless from the original VOB ... though I should check into that too.)


    Thanks again! I'm inching ever closer to my time-consuming goal.
    Quote Quote  
  4. Originally Posted by jephyork View Post
    It looks like if I set Streamclip to "DO NOT deinterlace", and set the output to 29.97fps, the result maintains the VOB's current interlacing -- just combining every two adjacent fields into a frame. I get three clear frames, then two interlaced (toothy) frames. Obviously the two toothy frames are a problem -- but their existence seems to indicate that Streamclip is now KEEPING and COMBINING both fields, not throwing one set away.

    So, first question: do I have that right?
    Not exactly. It's not combining anything since that's the way it's stored on the DVD (or whatever your source is), in the telltale 3 progressive/2 interlaced frame cycle.
    - Separate all the fields.
    - Delete the very first field, and the very last field, of the clip. (I'll add a junk frame at the start and end so I don't lose anything I need.)
    - Weave the fields back together.

    This should get me what I need. Right?
    Why waste all that time doing it manually when a good AviSynth IVTC will do it all for you, and automatically? For example, the AviSynth TIVTC of:

    TFM().TDecimate()

    should give you what you want (progressive and 23.976fps) with no additional intervention from you at all.

    Also: can it do this losslessly?
    No, but you're going to reencode anyway. You can do the IVTC and the upscaling all in the same script.

    And what format/codec would the original Streamclip file need to be, to work in AviSynth?
    Forget Streamclip. Open whatever your source is using the appropriate AviSynth source filter. What is it? A DVD?

    Far and away the best way to accomplish what you want is using a Windows based computer and AviSynth.
    Quote Quote  
  5. Member
    Join Date
    Mar 2017
    Location
    Salem, MA
    Search Comp PM
    Probably so, but I don't know how to use those things.

    I can't use an automated IVTC because my DVD source has broken cadences.
    Quote Quote  
  6. Originally Posted by jephyork View Post
    Probably so, but I don't know how to use those things.

    I can't use an automated IVTC because my DVD source has broken cadences.

    How broken ? To what extent and pattern ?

    99% of the time it doesn't really matter with broken cadences, because TIVTC is an adaptive field matcher. TFM matches all the fields in a range (they can be out of order, out of place, etc.. ) , TDecimate throws away the duplicate. You're still left with the original 24p film frames .

    Let me put it another way: if you manually specified which fields to match, and which to discard, and which to reconstruct, you'd basically get the same thing on most "broken" cadence sources. But this will save you hours or days of work. The common exceptions are some types of animation (which may require different types of treatment, but they can have different underlying frame rates to begin with) , and interlaced fades (which can be tricky with automatic treatment) . And if you have one of those tricky exceptions, post a sample and someone will help you with the proper settings to use (there are manual overrides possible too for TIVTC , but 99.999% of the time you don't need to use it)

    The fallback for TFM post processing is comb detection and a "dumb" deinterlace only for those frames after field matching that are detected as "combed" . But there are dozens of settings for TFM and TDecimate and you can customize for your taste. For example you could specify a higher quality interpolation like QTGMC for the combed frames, or maybe you want to adjust thresholds for comb detection, etc..

    You don't want to use mpegstreamclip for anything. Waste of time. And to get something out of mpegstreamclip it will be lossy unless you use a lossless YUV codec. If you are using avisynth you would use DGIndex on the original DVD . That's probably one of the big reasons why avisynth approach gives consistently better results - the source filter. It's indexed and frame accurate with non linear seeks. FFMpeg, despite using the same ported TIVTC code, has "slips" where it either matches wrong field, or discards wrong frame.
    Quote Quote  
  7. Member
    Join Date
    Mar 2017
    Location
    Salem, MA
    Search Comp PM
    Ah, nothing like coming to the Mac forum for Mac help with my Mac software, and being told to use a PC.

    Still -- thanks, I do appreciate the responses. Let me explain a little more about my project and my source material.

    The source is an action/adventure TV show from 20 years ago, now out on DVD. It was shot on film, but transferred to videotape and interlaced *before* it was edited. So every single time there's a cut, the cadence breaks. And because it's a fast-paced show, there are a LOT of cuts.

    There are also interlaced fades. There are VFX and title sequences that were rendered natively at 29.97. There are fight scenes that were sped up 2x, and now each field represents all that's left of its frame. It's a bit of a mess and some fields will just plain have to be interpolated into lossy frames. But most won't! That's why I wanted to do it manually -- sure, the process may take 100 times longer, but I'll get to examine each shot myself and see what's going on with it.

    My goal is to reconstruct the original 30-second promos for each episode, using my old VHS recordings as a guide. Meaning I'm not extracting a single 30-second clip from an episode -- I'm taking dozens of 1-to-3-second shots from all throughout the episode, and editing them into one clip. Then, once it's assembled, I plan to upscale it to 720p (or maybe 1080p). Why? Because! (It's a gift for a friend who complains bitterly that the show will never be released in HD.)

    My final framerate will not be 23.976 or even 29.97, it'll be 59.94 -- to accommodate all the weirdness from the VFX shots and the sped-up fight shots. I basically plan to replace each *field* on the old VHS copy with its full *frame*, and double the framerate.

    This may be weird and dumb and a lot of extra work, but this is a hobby for me and I think I'm going to enjoy it. I just need to figure out how to get at the fields, how to make them (wherever possible) into the frames they once were -- and hopefully how to do it losslessly, so the final upscale looks as good as it can.


    Poisondeathray, are you saying that Streamclip is lossy from the get-go? Earlier you seemed to imply the opposite.

    What does "use DGIndex on the original DVD" mean? It looks to be similar to Streamclip ... are you saying I should use it to transcode the VOB instead? (Hooray, more PC-only software to master!)


    I had been using: MacTheRipper to get the VOB off the DVD; Streamclip to transcode the VOB into an MPG; and Final Cut Pro to edit it.

    I'd still like to use Final Cut Pro to edit, because I'm familiar with it. (Is that possible?) But if you're saying that I should use other, PC-based programs to get the file into editable shape ... well, I'm afraid I'm somewhat lost now.

    Can you recommend a lossless workflow?
    Last edited by jephyork; 22nd Mar 2017 at 21:46.
    Quote Quote  
  8. There are also interlaced fades. There are VFX and title sequences that were rendered natively at 29.97. There are fight scenes that were sped up 2x, and now each field represents all that's left of its frame. It's a bit of a mess and some fields will just plain have to be interpolated into lossy frames. But most won't! That's why I wanted to do it manually -- sure, the process may take 100 times longer, but I'll get to examine each shot myself and see what's going on with it.
    Those interlaced fade sections can be tricky because even if you manually match, you often will get combing/artifacts anyways - and you have to post process with filters anyways unless you want to leave it looking crappy (again a job for avisynth or vapoursynth)

    Also, the single field interpolation is very poor in NLE's like FCP, Premiere etc.. , programs like mpegstreamclip . They just resize the field into a frame. Aliasing and jaggies



    My final framerate will not be 23.976 or even 29.97, it'll be 59.94 -- to accommodate all the weirdness from the VFX shots and the sped-up fight shots. I basically plan to replace each *field* on the old VHS copy with its full *frame*, and double the framerate.

    Yes, 59.94 is "safe" in that all common framerates (at least ones on DVD) can be stored in the 59.94 frames per second. For example 29.97 sections will have perfect duplicates. 23.976p sections will have 3:2 duplicates. Animation running at 12 fps unique content will have 5 frame duplicates, etc...

    In that 1% tricky category is the option to run VFR (variable frame rate). This means 23.976p sequences run 23.976p, 29.97p sequences run 29.97, 59.94 true interlaced content runs 59.94, etc... ie. the frame rate actually changes and you're left with unique frames instead of duplicate frames in sections .

    But when you view on a 60Hz display, VFR will look the same anyways. For example the 23.976 section will still have that 3:2 judder. The reason some people use VFR is because there are no hardcoded duplicates (slightly more efficient encoding wise because fewer frames are encoded, thus slightly higher quality at a given bitrate)




    Poisondeathray, are you saying that Streamclip is lossy from the get-go? Earlier you seemed to imply the opposite.
    I meant when you export it , it can become lossy depending on what you use. You can't get back the original MPEG2 DVD compression. It's decompressed / decoded into the uncompressed data when you process it in mpegstreamclip (or avisynth ). There is no way to field match or re-arrange fields and keep the same original compression losslessly. It has to be decompressed, recompressed. And that recompression stage can be lossy or lossless

    If you're famillar with audio, an analogy would be flac uses lossless compression, but something like mp3 is lossy, throwing away bits of data. Or in image processing BMP, TIFF, PNG are lossless, but something like JPG is lossy, throwing away bits of data

    I think you were concerned about deinterlacing progressive content. And that is avoidable quality loss.



    What does "use DGIndex on the original DVD" mean? It looks to be similar to Streamclip ... are you saying I should use it to transcode the VOB instead? (Hooray, more PC-only software to master!)
    It's just a utility, and the source filter MPEG2Source, is the most accurate , reliable way to load mpeg2 sources into avisynth




    I'd still like to use Final Cut Pro to edit, because I'm familiar with it. (Is that possible?) But if you're saying that I should use other, PC-based programs to get the file into editable shape ... well, I'm afraid I'm somewhat lost now.
    You can use FCP to edit, but it doesn't have any capability field match adaptively, only set pulldown cadence removal. Same with other software like after effects. They use set cadences patterns; or treat as interlaced. Ie. not adaptive like TIVTC, only set repeating patterns





    Can you recommend a lossless workflow?
    Operations like field matching are lossless, until you export. If you use uncompressed or lossless compression (huge filesizes) then it's "lossless". Most people will want reduced managable filesize for final distribution, thus lossy compression - so that's not lossless
    Quote Quote  
  9. Originally Posted by jephyork View Post
    Ah, nothing like coming to the Mac forum for Mac help with my Mac software, and being told to use a PC.
    For the kind of work you're attempting, there's not much in the way of alternatives, not if you want to do it right.
    I can't use an automated IVTC because my DVD source has broken cadences.
    Of course you can, or you can use as much manual intervention as you wish. For fields that can't be matched, the IVTC will either deinterlace or drop orphaned fields. Yes, it having been edited as video with no concern for the cadence presents problems, but yours isn't the first such instance. Music videos do that crap all the time.

    My final framerate will not be 23.976 or even 29.97, it'll be 59.94
    Is this the first you've mentioned that? So, bob it and be done with it. I expect that can be done on a Mac, but AviSynth has far and away the best deinterlacers and bobbers.

    What does "use DGIndex on the original DVD" mean?
    DGindex 'Indexes' the VOB so it can be used in an AviSynth script with the MPEG2Source filter to open the video in Virtual Dub or your encoder or whatever can read AviSynth scripts, including most players. Going the DGIndex route is the best and safest way to deal with DVD sources. It's all explained in the docs that accompany the DGMPGDec package. DGIndex encodes nothing - doesn't change the video in any way.

    Can you recommend a lossless workflow?
    If there are to be intermediate stages, then for each stage you save out your work as a lossless AVI using any one of a number of codecs, ones such as Lagarith or UTVideo Codec. Those AVIs can, in their turn, be opened in AviSynth scripts using the AviSource source filter. You can keep it lossless until the final encode. Nothing you've mentioned so far is very difficult to accomplish on a PC with even a rudimentary knowledge of AviSynth.

    Maybe you can stumble through it on a Mac, but if it entails trying to do what you were asking about in your first post, you'll quickly regret it. People doing this kind of work use PCs. But maybe pdr or someone else can recommend a Mac-only method that will give you something.
    Last edited by manono; 23rd Mar 2017 at 04:22.
    Quote Quote  
  10. One problem with avisynth is there is no nice GUI. So if you're putting a bunch of clips into a compliation/edit, it's more difficult to get timing etc.. A NLE like FCP/PP etc... will be 10x easier and faster to use for that part. Avspmod is sort of like a GUI, but a far cry from a proper NLE.

    If you were certain there were only 23.976p and 29.97p sections, what I would do in your shoes, is use TFM with clip2 specifying QTGMC. This will give you proper full resolution field matching and result in 29.97p . You can chose to manually decimate in FCP (by manually cutting out the selected 1 in 5 duplicate in the "23.976p in 29.97p" sections)... but TDecimate is smart and adaptive. Whereas a varying pattern will fool FCP/ AE etc..., it doesn't matter for TDecimate. eg. If the duplicate is 1st , or 5th in another section, 3rd in another etc.. it will pick the right one to discard. So for known 23.976 sections, I would also apply TDecimate. Thus, you would have various 23.976p and 29.97p clips, and import them into a 59.94p timeline in FCP to do the compliation. The reason for clip2 QTGMC, is instead of simple field resize (for only those sections that have a single field, or that have combing), you get higher quality interpolation for the reconstructed frame. To import into FCP, the "best" format is prores HQ. Technically it's not lossless, but it's probably close enough for most people.

    In theory, you could replace the avisynth part (without QTGMC) with ffmpeg osx binary with the field match (+/- decimate for 23.976 sections) mentioned above. Super simple and batch processing. But in reality the results are worse. I don't understand it because the code was ported directly from avisynth... I suspect part of the difference is from using a rock solid source filter like mpeg2source in avisynth; because when you use an "unreliable" source filter in avisynth but everything else the same - you can get buggy results too in avisynth
    Last edited by poisondeathray; 22nd Mar 2017 at 23:07.
    Quote Quote  
  11. Member
    Join Date
    Mar 2017
    Location
    Salem, MA
    Search Comp PM
    Well, I'm slowly getting VERY lost in this discussion.

    I appreciate all the replies and advice, but I can't follow the specifics of most of it anymore. Sentences like "bob it and be done with it" or "use TFM with clip2 specifying QTGMC" just go right over my head.

    Let me try to address what I can:

    One problem with avisynth is there is no nice GUI.
    Yeah -- I'm a big baby and I like a GUI for most everything. Command-line stuff and scripts stop me right in my tracks.

    Those interlaced fade sections can be tricky because even if you manually match, you often will get combing/artifacts anyways
    I'd just interpolate those fields out into full frames. Seems easiest. Image loss but a smoother fade, no?

    Also, the single field interpolation is very poor in NLE's like FCP, Premiere etc.. , programs like mpegstreamclip . They just resize the field into a frame
    Good to know. I wasn't planning to use FCP to do the interpolation -- I had assumed there'd be some whiz-bang piece of other software that could do a better job. Probably PC-only, no doubt.

    I like the advice on variable framerate to save space -- but these are 30-second clips and I honestly don't care if they're huge. At this point in my "career" I think I should keep it simple, and stick with duplicated frames and a bigger file.

    On working "losslessly" -- yeah, I'm aware that Streamclip and other programs "uncompress" the DVD compression into a giant file, and then at the end I'd have to "recompress" it in a different way. (Or just keep the huge file, I guess.) I'm just paranoid that with each change I make to the frames as I go along -- combining the fields, converting to a different codec, etc -- I somehow degrade the image data in an avoidable way.

    When I ask for "lossless" ways to get the VOB through Streamclip or readable by AviSynth or "lossless" ways to weave fields together -- I'm just hoping to avoid degrading the frames. (Like interpolation does, as far as my understanding goes.)


    I'm going to download Wine and AviSynth and do some reading and poking at them, see if I can even get them to work. But I'd still like to find a way to "stumble through it on a Mac", as manono says. For me, personally, it seems like it'll be a lot more comfortable to go slow using tools and computers that I already know how to use -- rather than starting completely over from scratch and teaching myself an entire new operating system, just so I can delete duplicate frames with greater ease.

    Thanks again to both of you for your advice and patience. If you've got more thoughts I'd love to hear them.
    Quote Quote  
  12. To bob something in this context just means to turn the individual fields into frames. Or, in the case of your DVD, turn 29.97 interlaced frames per second into 59.94 progressive frames per second. It's what you said you wanted to do. And I responded by saying AviSynth has the best bobbers. I know nothing about Macs or Mac software but the QTGMC bobber pdr mentioned beats the expensive software you'd prefer to use hands down.

    You've already spent a ton of time on the project so I'm assuming you want good results from the time you've invested. Bobbing it is a valid way to handle a difficult source and QTGMC is the best bobber, bar none.
    Quote Quote  
  13. Originally Posted by jephyork View Post

    Those interlaced fade sections can be tricky because even if you manually match, you often will get combing/artifacts anyways
    I'd just interpolate those fields out into full frames. Seems easiest. Image loss but a smoother fade, no?
    It depends on the type of "interlaced fade" (there are slightly different causes). But regardless of the underlying cause, they almost always "mess" up the automatic algorithms. The most common way is just "brute force" and treating the residual combing with a filter called vinverse() . There is actually a better solution in vapoursynth but it's more complicated . Eitherway , I would treat those sections differently .

    The other tricky scenario is where you have graphics , maybe credits sometimes that run at a different rate. So the underlying content might be "film" rate 23.976, but you might have interlaced graphics, titles, animations running overlayed on top of the base image at either 29.97 or 59.94 (when bobbed) . There are special ways to deal with those as well

    Also, the single field interpolation is very poor in NLE's like FCP, Premiere etc.. , programs like mpegstreamclip . They just resize the field into a frame
    Good to know. I wasn't planning to use FCP to do the interpolation -- I had assumed there'd be some whiz-bang piece of other software that could do a better job. Probably PC-only, no doubt.
    Yes....it's called avisynth . But vapoursynth can do it too and you can run it natively on a mac.



    On working "losslessly" -- yeah, I'm aware that Streamclip and other programs "uncompress" the DVD compression into a giant file, and then at the end I'd have to "recompress" it in a different way. (Or just keep the huge file, I guess.) I'm just paranoid that with each change I make to the frames as I go along -- combining the fields, converting to a different codec, etc -- I somehow degrade the image data in an avoidable way.

    When I ask for "lossless" ways to get the VOB through Streamclip or readable by AviSynth or "lossless" ways to weave fields together -- I'm just hoping to avoid degrading the frames. (Like interpolation does, as far as my understanding goes.)
    There is no way in streamclip or fcp to field match. (Well FCP can , but only in a fixed pattern. As soon as you feed an irregular pattern it screws up - i.e. non adaptive)

    Streamclip can't even do that. It can only deinterlace (which degrades progressive content as you've seen) , or pass through as original interlaced fields (which means it's useless for this, because why use streamclip at all, just use the original)



    I'm going to download Wine and AviSynth and do some reading and poking at them, see if I can even get them to work. But I'd still like to find a way to "stumble through it on a Mac", as manono says. For me, personally, it seems like it'll be a lot more comfortable to go slow using tools and computers that I already know how to use -- rather than starting completely over from scratch and teaching myself an entire new operating system, just so I can delete duplicate frames with greater ease.

    Thanks again to both of you for your advice and patience. If you've got more thoughts I'd love to hear them.
    You can try avidemux (GUI), with the filters mentioned above. If you use the first filter only, that's the "decomb" method of field matching. But avidemux is unreliable IMO and crashes (often) .

    You can also completely manually specify fields etc.... in avisynth (I don't see any benefit from what you've described; the semi-automated approach will give the same results unless you have a very bad source, things like heavy noise, misaligned fields...), but there is a utility called yatta that can assist with manual matching. But these are for windows; you can use vapoursynth natively on a mac but it's script based and even more complicated than avisynth

    I don't know of any GUI mac based utility where you can field match with manual override or control. There is actually none that can adaptively field match in the first place, with the exception of the decomb ported filters in avidemux
    Quote Quote  
  14. Member
    Join Date
    Mar 2017
    Location
    Salem, MA
    Search Comp PM
    I haven't vanished. Been reading up on AviSynth, doing my research. It's a lot to take in.

    Is there an "AviSynth for complete beginners" forum somewhere out there, where I could ask my various basic/dumb questions and get help writing scripts?
    Quote Quote  
  15. The anime music videos site has some well written guides to using AviSynth:

    http://www.animemusicvideos.org/guides/avtech31/

    The three docs included in the DGMPGDec package are an excellent introduction to preparing DVD video for use in AviSynth scripts.

    The AviSynth site has some pages dedicated to people just starting out:

    http://avisynth.nl/index.php/Main_Page

    Doom9 has a whole forum dedicated to AviSynth and others to some related subjects. It's well worth reading there to learn, but it's not really geared for people asking really basic questions:

    https://forum.doom9.org/forumdisplay.php?f=33

    Your best place for asking rookie questions is here at videohelp.com
    Quote Quote  



Similar Threads

Visit our sponsor! Try DVDFab and backup Blu-rays!