VideoHelp Forum

Our website is made possible by displaying online advertisements to our visitors. Consider supporting us by disable your adblocker or try DVDFab and copy, convert or make Blu-rays and DVDs! :)
+ Reply to Thread
Page 3 of 3
FirstFirst 1 2 3
Results 61 to 89 of 89
Thread
  1. StaxRip is good for beginners, certainly, but you don't have to use it to encode anything. A relatively easy alternative might be VirtualDub2 through which you can make your test encodes or your full encodes. Open the AviSynth script, select whatever codec you want from the menu in the Video->Compression area (x264 8-bit?) and Configure it (Default CRF 18, Tuning=Film?), be careful of SAR). Then go File->Save As and choose the container (MP4, MKV, whatever). If you want audio, choose and configure it in the Audio section.
    Quote Quote  
  2. Member
    Join Date
    Dec 2018
    Location
    Spain
    Search Comp PM
    Originally Posted by manono View Post
    StaxRip is good for beginners, certainly, but you don't have to use it to encode anything.
    The main reason for me to try it was that I could use QTGMC without dealing with AviSynth at all... but that ship has sailed.

    Originally Posted by manono View Post
    A relatively easy alternative might be VirtualDub2 through which you can make your test encodes or your full encodes. Open the AviSynth script, select whatever codec you want from the menu in the Video->Compression area (x264 8-bit?) and Configure it (Default CRF 18, Tuning=Film?), be careful of SAR). Then go File->Save As and choose the container (MP4, MKV, whatever). If you want audio, choose and configure it in the Audio section.
    Thanks for your suggestion, I'm already encoding with it to see how it goes. I have a question about the audio file I generated with DGIndex: the one for the complete video says DELAY 0ms in its filename. The one from the short clip I made says DELAY -67ms. Why is that? Does it have something to do with how/where I made the cut? The audio does indeed come out of sync (clearly more than 67ms, otherwise I wouldn't have noticed), but I haven't found a way to correct it yet. I also noticed the first few frames of my short clip seem odd, a couple of them seem to be repeated. I suspect both things are related. Not that I care too much about audio for my test clips, just curious.

    I only recently became aware that x264 has tunes. I've never used one, but if you recommend to use film, I'll give it a try. Would that be your go-to tune except for specific cases (animation, any other case covered by the other tunes) or is there any case where you'd advice against it?
    Last edited by MGRV; 19th Dec 2018 at 15:23.
    Quote Quote  
  3. Originally Posted by manono View Post
    For example, by adding:

    Trim(1000,2000)

    at the bottom of your script, you'll encode frames 1001-2001.
    Also note that you're not limited to a single Trim. You can use it to select multiple segments.:

    Code:
    Trim(1000, 2000) + Trim(11000, 12000) + Trim(25000, 26000)
    The plus sign appends the three segments into a single clip.
    Quote Quote  
  4. Originally Posted by MGRV View Post
    The one from the short clip I made says DELAY -67ms. Why is that?
    Beats me, except you always get delays with cut VOBs. And I always take them into account when working with the audio or muxing. You can do away with the delay entirely by figuring out how much it is (you said that for the cut file it was obviously larger than -67ms) and then using DelayCut to get rid of it.

    I've never used one, but if you recommend to use film, I'll give it a try.
    It was a suggestion more than a recommendation. Sounds good for a movie, though. Others know more than I about tuning. Me, I use RipBot64 for my MP4 encodes. It's the same idea as VDub2. Open an AviSynth script, open audio, and after that you can tweak as much or as little as you like before encoding.
    Quote Quote  
  5. Member
    Join Date
    Dec 2018
    Location
    Spain
    Search Comp PM
    Originally Posted by jagabo View Post
    Also note that you're not limited to a single Trim. You can use it to select multiple segments.
    Yes, I've been playing around with that and there's already a new question on its way (about making a video out of two different sources ), but I think I'd better hold that off until I'm done with this project. Better not to sink my teeth into too many things at once. Should I post it here in the newbies section or in Editing, rather? I'm not under the delusion that I've somehow 'graduated', they will still be newbie editing questions!

    Originally Posted by manono View Post
    Beats me, except you always get delays with cut VOBs. And I always take them into account when working with the audio or muxing. You can do away with the delay entirely by figuring out how much it is (you said that for the cut file it was obviously larger than -67ms) and then using DelayCut to get rid of it.
    So, as long as I'm taking whole streams from the VOB files with no cuts done, should I get theoretically zero delay every time? Do other formats have this problem, or is it just VOBs?

    Originally Posted by manono View Post
    It was a suggestion more than a recommendation. Sounds good for a movie, though. Others know more than I about tuning.
    I guess I've always been a bit wary of those kinds of preset tunes. Not knowing well what they do, I figured they could be too intrusive of the original source. But I've noticed 'film' in particular is broadly used. I'll have to give it a try and compare results. Probably at a latter stage, as I don't want to throw more variables into the mix, and my eyes are claiming for mercy trying to cope with all the QTGMC testing.


    PS: I realized I did something pretty stupid: I used VirtualDub2 per manono's suggestion, but used it to encode a script that pointed to my already cut "sample D2V" as a source instead of the (trimmed) original D2V. Now I correctly referenced the original D2V, but using the demuxed audio won't work either, as the audio runs from the start while my video is trimmed. Maybe the AudioDub function helps here, but I don't really need to complicate things too much, I can do away with audio for these short clips.
    Last edited by MGRV; 20th Dec 2018 at 02:40.
    Quote Quote  
  6. Member
    Join Date
    May 2014
    Location
    Memphis TN, US
    Search PM
    Originally Posted by MGRV View Post
    PS: I realized I did something pretty stupid: I used VirtualDub2 per manono's suggestion, but used it to encode a script that pointed to my already cut "sample D2V" as a source instead of the (trimmed) original D2V. Now I correctly referenced the original D2V, but using the demuxed audio won't work either, as the audio runs from the start while my video is trimmed. Maybe the AudioDub function helps here, but I don't really need to complicate things too much, I can do away with audio for these short clips.
    Open the audio and video from the d2v and ac3 files as shown earlier, and use AudioDub() to remux the two together. You do this in a script to be sure that when you make later cuts or trims, the audio and video gets cut and trimmed at the same time and stays in sync.
    - My sister Ann's brother
    Quote Quote  
  7. Originally Posted by MGRV View Post
    there's already a new question on its way (about making a video out of two different sources )
    That was basically covered at the end of post #40

    https://forum.videohelp.com/threads/391244-Converting-DVDs-Suggestions-for-an-absolute...e2#post2536895

    With a minor difference: instead of interleaving, add them:

    Code:
    v1 = WhateverSource("file1.ext")
    v2 = WhateverSource("file2.ext")
    v1 + v2
    You may filter them before and/or after joining them.

    Originally Posted by MGRV View Post
    So, as long as I'm taking whole streams from the VOB files with no cuts done, should I get theoretically zero delay every time?
    No. More often than not there are audio delays in VOB files, whether you cut them or not.

    Originally Posted by MGRV View Post
    Do other formats have this problem, or is it just VOBs?
    Most containers support audio delays. So you may see them anywhere.

    Originally Posted by MGRV View Post
    I guess I've always been a bit wary of those kinds of preset tunes. Not knowing well what they do
    With x264 you can see the settings that were used with MediaInfo. So you could encode a short clip with different tunings and compare the settings.

    Code:
    preset=slow, tune=none, crf=18: cabac=1 / ref=5 / deblock=1:0:0   / analyse=0x3:0x113 / me=hex / subme=8 / psy=1 / fade_compensate=0.00 / psy_rd=1.00:0.00 / mixed_ref=1 / me_range=16 / chroma_me=1 / trellis=2 / 8x8dct=1 / cqm=0 / deadzone=21,11 / fast_pskip=1 / chroma_qp_offset=-2 / threads=6 / lookahead_threads=1 / sliced_threads=0 / nr=0 / decimate=1 / interlaced=0 / bluray_compat=0 / constrained_intra=0 / bframes=3 / b_pyramid=2 / b_adapt=1 / b_bias=0 / direct=3 / weightb=1 / open_gop=0 / weightp=2 / keyint=250 / keyint_min=23 / scenecut=40 / intra_refresh=0 / rc_lookahead=50 / rc=crf / mbtree=1 / crf=18.0000 / qcomp=0.60 / qpmin=0 / qpmax=69 / qpstep=4 / ip_ratio=1.40 / aq=1:1.00
    preset=slow, tune=film, crf=18: cabac=1 / ref=5 / deblock=1:-1:-1 / analyse=0x3:0x113 / me=hex / subme=8 / psy=1 / fade_compensate=0.00 / psy_rd=1.00:0.15 / mixed_ref=1 / me_range=16 / chroma_me=1 / trellis=2 / 8x8dct=1 / cqm=0 / deadzone=21,11 / fast_pskip=1 / chroma_qp_offset=-3 / threads=6 / lookahead_threads=1 / sliced_threads=0 / nr=0 / decimate=1 / interlaced=0 / bluray_compat=0 / constrained_intra=0 / bframes=3 / b_pyramid=2 / b_adapt=1 / b_bias=0 / direct=3 / weightb=1 / open_gop=0 / weightp=2 / keyint=250 / keyint_min=23 / scenecut=40 / intra_refresh=0 / rc_lookahead=50 / rc=crf / mbtree=1 / crf=18.0000 / qcomp=0.60 / qpmin=0 / qpmax=69 / qpstep=4 / ip_ratio=1.40 / aq=1:1.00
    Quote Quote  
  8. Member
    Join Date
    Dec 2018
    Location
    Spain
    Search Comp PM
    Originally Posted by jagabo View Post
    That was basically covered at the end of post #40
    Oh, I got that, I was getting into "ambitious newbie" mode. Ok, here it goes: I have 2 videos of the same footage. One has much better quality, the other one is more complete. A long time ago, I tried merging both videos into one, using the complete version to 'patch' the missing parts of the better-quality version. I was using some freeware GUI I can't recall and didn't really bother to learn much about what I was doing. I didn't really need it at all, I tried it just for fun and got pretty carried away. That was until I tried the first encodes and the results were an awful blocky mess. Now, having seen just the tip of the iceberg, I see how many steps of the process I could have screwed up to get such poor results.

    Just for kicks, I tried a combination of trimming and appending which works as expected, although the audio gets out of sync.

    Code:
    clip1 = AviSource("better_video.avi")
    clip2 = AviSource("complete_video.avi")
    
    clip1a = Trim(clip1, 0, 1000)
    clip1b = Trim(clip1, 1001, 2000)
    clip1c = Trim(clip1, 2001, 0)
    clip2a = Trim(clip2, 1001, 1500)
    clip2b = Trim(clip2, 2501, 3000)
    
    clip1a + clip2a + clip1b + clip2b + clip1c
    So I know it can be done, my question was going to be more about whether this can be done without quality loss of at least the better source (I could re-encode the other one to match its resolution and whatever else need to be the same).


    Originally Posted by jagabo View Post
    More often than not there are audio delays in VOB files, whether you cut them or not.
    I guess I was lucky to get 0 delay on my first attempt. I have a few of these to convert, so I'll eventually have to deal with that. I'll check DelayCut, though I'm a bit confused about this whole deal. I suppose because there are different moments in the process where one can join audio+video, there are also different ways to deal with sync problems.



    Originally Posted by LMotlow View Post
    Open the audio and video from the d2v and ac3 files as shown earlier, and use AudioDub() to remux the two together. You do this in a script to be sure that when you make later cuts or trims, the audio and video gets cut and trimmed at the same time and stays in sync.
    Is there any source filter for audio that comes with AviSynth, or should I download one in order to load my AC3 file?
    Quote Quote  
  9. Originally Posted by MGRV View Post
    So, as long as I'm taking whole streams from the VOB files with no cuts done, should I get theoretically zero delay every time?
    The only time you'll usually get no delay is with the first VOB of a video. And not always even the. All the rest, whether at the beginning of a subsequent VOB or if a VOB is cut will have a delay.

    I've been playing around with that and there's already a new question on its way (about making a video out of two different sources ), but I think I'd better hold that off until I'm done with this project.
    I do that with some regularity and do it this way, saving keeping track of all the trims. It requires both sources have the same frame count and resolution.

    A=MPEG2Source("Movie.d2v")
    B=MPEG2Source("BetterMovie.d2v")
    ReplaceFramesSimple(A,B,Mappings="[1000 2000] [10000 12000] 20000 25555] ")


    ReplaceFramesSimple is a part of stickboy's RemapFrames. The frame numbers within the brackets are the start and end frame numbers for the replacement. Movie gets replaced with BetterMovie in those sections.

    So I know it can be done, my question was going to be more about whether this can be done without quality loss of at least the better source
    You're using AviSynth, so everything gets reencoded, if that's what you're asking.

    Is there any source filter for audio that comes with AviSynth, or should I download one in order to load my AC3 file?
    You'll need NicAudio for that.
    Quote Quote  
  10. Originally Posted by MGRV View Post
    editing in AviSynth... I know it can be done, my question was going to be more about whether this can be done without quality loss
    Only uncompressed video and audio pass through AviSynth. If you (re)compress with a lossy codec you will lose quality. I you save uncompressed or losslessly compressed you may not lose quality but you will get very large files. You have to decide if the benefits of using AviSynth outweigh the drawbacks.
    Quote Quote  
  11. Member
    Join Date
    Dec 2018
    Location
    Spain
    Search Comp PM
    Originally Posted by manono View Post
    The only time you'll usually get no delay is with the first VOB of a video. And not always even the. All the rest, whether at the beginning of a subsequent VOB or if a VOB is cut will have a delay.
    I'll have to add DelayCut to my army of tools, then. I can use the out-of-sync file in PotPlayer to figure how much I need to cut/delay.

    You'll need NicAudio for that.
    I tried downloading it from the Codeplex site but couldn't find the .dll in the zip file [found it!]. I'll look around for a site that looks safe. In the meantime, I tried LWlibavAudioSource but either it doesn't work with AC3 or I made some mistake.


    Originally Posted by jagabo View Post
    Only uncompressed video and audio pass through AviSynth. If you (re)compress with a lossy codec you will lose quality. I you save uncompressed or losslessly compressed you may not lose quality but you will get very large files. You have to decide if the benefits of using AviSynth outweigh the drawbacks.
    See, that was me being a newbie. I sort of knew that, but yet the salad in my brain made me get it all wrong. Now, back to being ambitious...

    Is there a way to do it losslessly (outside AviSynth)? I understood that's what ffmpeg does with concat and using copy, but I'm not sure about anything anymore! I mean, having two videos that share the same specs and joining different segments of them into one single video, without having to decompress them. If that's the case, what I was thinking is I could re-encode only the complete video (worse quality, could use some filtering) and encode it so that it matches the better one before doing the joins.
    Last edited by MGRV; 20th Dec 2018 at 13:42. Reason: Found NicAudio.dll !
    Quote Quote  
  12. Originally Posted by MGRV View Post
    Is there a way to do it losslessly (outside AviSynth)?
    If the goal is another DVD, you can cut out the 'good' pieces (using DGIndex), reencode the 'bad' pieces to M2V and then join them one after the other using the 'Add' button in Muxman. With a lot of joining to do, it hardly seems worth it to me.

    I checked that NicAudio link and couldn't figure it out either. I'll enclose the DLL and TXT file.
    Image Attached Files
    Quote Quote  
  13. Member
    Join Date
    Dec 2018
    Location
    Spain
    Search Comp PM
    Originally Posted by manono View Post
    If the goal is another DVD, you can cut out the 'good' pieces (using DGIndex), reencode the 'bad' pieces to M2V and then join them one after the other using the 'Add' button in Muxman.
    No, sorry I didn't make that clear. I'm pretty sure the videos were an mp4 (the 'quality' source) and an avi (the 'complete' source). My output file would be... whatever format I can play, though if there's any chance to avoid encoding the 'quality' source, I guess the target format would be the same (mp4). Maybe what I'm saying is not even possible, but my idea would be to decompress/filter/reencode the complete source, in order to make it compatible with the 'quality' source for joining (plus applying my vastly new knowledge about filtering, of course).

    I checked that NicAudio link and couldn't figure it out either. I'll enclose the DLL and TXT file.
    Thanks, I found it at http://avisynth.nl/index.php/NicAudio under 'Archived Downloads', I missed it the first time. I checked just in case and it seems yours is the same as the one I downloaded. I also tried and made it work with AudioDub, finally!
    Quote Quote  
  14. AVIs can be cut at keyframes without reencoding. That's by using Direct Stream Copy in VDub. So, you reencode the complete source to make it compatible with the quality source, Then you start direct stream copying the the various parts and eventually append all the pieces.
    Quote Quote  
  15. Originally Posted by MGRV View Post
    I tried LWlibavAudioSource but either it doesn't work with AC3 or I made some mistake.
    It works for me. I verified both 2.0 and 5.1 channel AC3 files.

    Code:
    V = Mpeg2Source("filename.d2v")
    A = LWlibavAudioSource("filename.ac3")
    AudioDub (V, A)
    Maybe your editor didn't like 6 channel PCM audio?
    Quote Quote  
  16. Member
    Join Date
    Dec 2018
    Location
    Spain
    Search Comp PM
    Originally Posted by manono View Post
    AVIs can be cut at keyframes without reencoding.
    Maybe I'm missing something. I'm barely starting to understand what VideoDub does and does not, and I know just enough about video formats to understand they are containers holding video and audio streams, but that's it. So, my apologies in advance for any dumb question. Shouldn't I be cutting two mp4s eventually (the so far untouched 'good source' and the AVI-to-MP4 reencoded 'complete source')? Can mp4s be cut at keyframes without reencoding as well?

    That's by using Direct Stream Copy in VDub.
    Noticing the Direct Stream Copy is what led me to ask if the 'slicing & joining' process could be done losslessly (totally forgetting I was doing my trims and appendages in an AviSynth script).

    Then you start direct stream copying the the various parts and eventually append all the pieces.
    How would this process go? Load one clip, select each part that makes it to the 'final cut' and direct stream copy to a separate mp4 file. Do the same with the second clip. Then append all of them, also in VideoDub?


    Originally Posted by jagabo View Post
    Originally Posted by MGRV View Post
    I tried LWlibavAudioSource but either it doesn't work with AC3 or I made some mistake.
    It works for me. I verified both 2.0 and 5.1 channel AC3 files.

    Maybe your editor didn't like 6 channel PCM audio?
    It didn't like me making mistakes.
    Quote Quote  
  17. VirtualDub's Direct Stream Copy is designed for use with AVI files. It can Direct Stream Copy AVS scripts but remember that it receives uncompressed video from an AVS script, not the original compressed video. In theory you may be able to remux your MP4 sources (depending on codecs, must be constant frame rate, etc.) into an AVI container and use Virtual Dub to cut/paste in Direct Stream Copy mode.

    Keep in mind that key frames may be hundreds of frames apart with long GOP codecs. So the granularity of your cuts won't be great.

    There is a class of "smart" editors that can cut frame accurately and only reencode cut GOPs. VideoRedo, for example, can do this with MPEG 2 and h.264 video (in several containers). I believe Vegas can do it with MPEG 2 video.
    Quote Quote  
  18. Member
    Join Date
    Dec 2018
    Location
    Spain
    Search Comp PM
    Originally Posted by jagabo View Post
    VirtualDub's Direct Stream Copy is designed for use with AVI files.
    Ah, I had no idea. I saw I could open MP4 files, didn't see Direct Stream Copy become grayed-out or anything and assumed it could be done. I tried it myself before asking what the problem was and saw I got an error that it can't be done with that video stream. So, Direct Stream Copy only works for AVI files. Duly noted.

    In theory you may be able to remux your MP4 sources (depending on codecs, must be constant frame rate, etc.)
    Yeah, I'm not clear on what exactly needs to match, but I guess the only way would be to experiment and see what happens. I only thought 'well, they have to match in resolution, DAR, etc' but I'm realizing there's way more to it. I only recently learned VFR was a thing.

    Keep in mind that key frames may be hundreds of frames apart with long GOP codecs. So the granularity of your cuts won't be great.
    I'll need to read more about that. I saw it mentioned while reading through different subjects in this forum and elsewhere, but I never understood it very well. I think I get the gist of it. So, depending on the codecs my sources were compressed with, I may not be able to do accurate cuts because I-frames (or keyframes, which I think means the same in this context?) will be too far apart from each other. I opened a random low-quality mp4 I had lying around and browsed through its frames in VideoDub. The second K frame was frame 60 (00:00:2.400 into the video). I checked with MediaInfo and it says GOP: M=1, N=60 (I couldn't quite understand what the M means, but I guess N is the GOP size I should care about). If this was my video, that would mean I can only place my edits every 2.400 seconds. Correct? Well, and as I think of it, there's no way the keyframes for both sources will match. So it's worse than I thought.

    There is a class of "smart" editors that can cut frame accurately and only reencode cut GOPs. VideoRedo, for example, can do this with MPEG 2 and h.264 video (in several containers). I believe Vegas can do it with MPEG 2 video.
    That's probably close to what I used. I don't remember if it was freeware or a trial version. [EDIT: Probably not. I think the one I used was a regular editor where you necessarily have to reencode the output. I just recently read about those 'smart editors' and I got confused.] It seemed to do what I needed, but I probably screwed up at the time of encoding. That is, assuming the software was correctly taking care of all these other things I was gleefully ignoring. Is there any argument to be made against such an editor in this case? If it only re-encodes the GOP in the bits it needs to 'stitch' together, it wouldn't be much of a drawback.
    Last edited by MGRV; 21st Dec 2018 at 10:57.
    Quote Quote  
  19. Originally Posted by MGRV View Post
    key frames... I'll need to read more about that.
    High compression codecs get a large part of their compression by not repeating parts of frames that are similar to other frames. Key frames (I frames, VirtualDub calls them K) are encoded as complete pictures, much like a JPEG image. Forward Predicted frames (P frames) reference prior frames (blocks of pixels are copied from prior frames to the P frame). Bidirectional predicted frames (B frames) are similar to P frames but they can copy from both frames before and/or after the B frame.

    A Group Of Pictures (GOP) is the I frame and its P and B frames, up to the next I frame. Since the B and P frames reference other frames, and ultimately the I frame, you can't remove the I frame without corrupting the other frames. So simple cut/paste editors that don't reencode the video have to cut-in on I frames (ie, any part you keep must start with an I frame). Some let you cut-out on P frames.

    Originally Posted by MGRV View Post
    Well, and as I think of it, there's no way the keyframes for both sources will match. So it's worse than I thought.
    Yes. Most encoders will start a new GOP at scene changes (or other large picture changes). So the configured GOP size is really the maximum GOP size, some GOPs will be smaller. It's not likely the key frames of two different encodings (done by different people with different sources) will have matching key frames everywhere. And, of course, to join parts of two different videos together most of the other properties must match. The frame rate, the frame size, the codecs, etc.
    Quote Quote  
  20. Member
    Join Date
    Dec 2018
    Location
    Spain
    Search Comp PM
    Originally Posted by jagabo View Post
    Originally Posted by MGRV View Post
    So it's worse than I thought.
    Yes.
    I may give VideoRedo a go sometime in the future. So, what it does is: decompress/reencode only the GOPs affected by my cuts, leave the rest untouched? As long as I provide it with workable sources (i.e. sharing certain specs). I thought it would be as easy as looking at the 'good' file with MediaInfo and trying to replicate the settings when encoding the other one, but I suspect it may be more than I can chew (sorry, I won't get rid of the tooth references until I'm done with these DVDs).
    Quote Quote  
  21. Member
    Join Date
    Dec 2018
    Location
    Spain
    Search Comp PM
    So, back to my original task. I'm considering sticking with QTGMC's fast preset, but I want to make sure it does enough getting rid of interlacing artifacts. What type of segments should I pay close attention to? I imagine: fast motion, camera pans, straight lines... Anything I'm missing?

    It's not like I'm noticing a huge loss of detail with slower presets. My eyes aren't trained to know what to look for, so I really don't know. It just so happens that I was expecting to see the infamous flying tooth and could hardly notice it in my first attempts with slower presets (which turn the 'tooth pixels' a bit darker, making them harder to spot against the background at such speed). It became obvious to me because I'm familiar with the footage and knew what to expect at that moment (how could I forget?), otherwise it would have passed me by entirely. But I took it as an indicator (maybe not an accurate one) of how the processing could affect the output elsewhere.

    Well, I'm trying to convince you that I'm a sane person not obsessed at all with the tooth, but who am I kidding? It was the first-ever televised bout in which the very first kick caused a tooth to fly. It's a dental piece of history. But if blending it a bit is what it takes to get rid of deinterlacing artifacts for good, there's always the slow-motion replay to fall to.
    Quote Quote  
  22. Originally Posted by MGRV View Post
    ...but I want to make sure it does enough getting rid of interlacing artifacts.
    Bobbers - almost by definition - get rid of all interlacing. The fields which they turn into frames - almost by definition - are progressive and interlace-free to begin with. If - after having used QTGMC or other bobbers - you think you still see interlacing, it's almost guaranteed to be something else, aliasing most likely. And QTGMC, because of the smoothing it does, is better about leaving less aliasing behind than something like a doublerate Yadif.

    What type of segments should I pay close attention to?
    Diagonal lines.
    Quote Quote  
  23. Member
    Join Date
    Dec 2018
    Location
    Spain
    Search Comp PM
    Originally Posted by manono View Post

    Bobbers - almost by definition - get rid of all interlacing. The fields which they turn into frames - almost by definition - are progressive and interlace-free to begin with. If - after having used QTGMC or other bobbers - you think you still see interlacing, it's almost guaranteed to be something else, aliasing most likely. And QTGMC, because of the smoothing it does, is better about leaving less aliasing behind than something like a doublerate Yadif.
    What I meant to say is that by using QTGCM's fast preset, I suspect there must be some risk of it 'not doing enough' to smooth out whatever it needs to (as opposed to a slower setting). I said 'interlacing artifacts' for lack of a better term, I wasn't referring to any particular type of artifact, though I did spot something (which is probably what is called aliasing). I'll upload the same frame from two different tests I did. The settings for them were something like this:

    QTGMC.png = QTGMC slow preset (maybe medium, can't recall)
    QTGMC_fast.png = QTGMC( Preset="fast", SourceMatch=3, Lossless=1, Sharpness=0.4, TR2=2)

    If you look at the fast QTGMC frame, the guy's shoulder has this serrated look (I'm struggling not to say sawtooth) while the other one looks much smoother.

    The settings for the fast one probably make no sense (I was just shooting in the dark), so it probably has to do more with that than with the fast processing. I'm trying just QTGMC(preset="fast") on VirtualDub now and it doesn't look nearly as bad.
    Image Attached Thumbnails Click image for larger version

Name:	QTGMC.png
Views:	9
Size:	387.4 KB
ID:	47551  

    Click image for larger version

Name:	QTGMC_fast.png
Views:	9
Size:	394.6 KB
ID:	47552  

    Last edited by MGRV; 21st Dec 2018 at 19:38.
    Quote Quote  
  24. Nearly horizontal edges and lines. Closely spaced (nearly) horizontal lines (like horizontal blinds). Examine such details while in motion. Not necessarily a lot of motion but with small motions. Watch for aliasing artifacts, buzzing edges, and moire artifacts.

    Also you often find that the two fields that make up a frame don't align perfectly horizontally. After a simple bob they will shake left/right. A smart bob might leave some comb artifacts (serrated vertical edges). The two fields may have slightly different intensities or colors. A simple bob will flicker. Smarter bobs may leave comb like artifacts because alternate scan lines have different intensities/colors. These types of problems can occur on still portions of the image as well as motion portions.
    Last edited by jagabo; 21st Dec 2018 at 19:39.
    Quote Quote  
  25. From the QTGMC Doc:

    The SourceMatch setting sets the mode: 0=off (standard [Q]TGMC algorithm), 1,2,3 for progressively more accurate but slower processing. Modes 2 & 3 restore almost exact source detail but are sensitive to noise & introduce occasional aliasing (mode 3 is less affected). Mode 1 is a more conservative halfway stage that rarely introduces artefacts.
    Quote Quote  
  26. Member
    Join Date
    Dec 2018
    Location
    Spain
    Search Comp PM
    Thank you both.

    Indeed, the aliasing was caused by my own contributions and not by fast processing. I was pretty sure I'd spotted something similar with the fast preset and no extra settings, but I don't see it now. I've done so many tests that it's likely I got them mixed up. So I guess it all boils down to this:

    Originally Posted by MGRV View Post
    I'm a bit biased against fast settings, as they're usually supposed to yield lower quality results
    I've noticed different results with the fast preset, but I'm not sure it makes for a 'lesser-quality' watching experience. I see less overall blurring. I could also see in frame-by-frame comparison that sometimes the slow preset got better results capturing certain objects, especially in the background. But I could swear I observed (less frequently) the exact opposite in other frames. And this is frame-by-frame, so I've yet to figure whether this eventual loss in quality is something that can be noticed watching at regular speed. I've incoporated too much new information and haven't been systematic about it, so I need to be patient and keep testing. But my aversion to the fast preset remains in the superstitious realm (as in, not grounded yet on factual evidence) though it's a reasonable assumption that slower is better (there's a reason why the slowest option is usually called placebo). I know from what was explained here and what I read that there are specific differences in the settings for each 'speed' and those values can be verified. But I haven't seen any problems with the fast preset when I watch the video (I thought I had, turns out I need to check again). I don't know what works and what doesn't, but I know perception can be very tricky. So I want to avoid being temporarily blindsided by the novelty of how good the video looks in general compared to other deinterlacing options, sticking to those settings and later finding out that the many DVDs I already encoded could have been improved substantially by a slower preset or some other adjustment.
    Quote Quote  
  27. Member
    Join Date
    Dec 2018
    Location
    Spain
    Search Comp PM
    I have a question about SAR (or PAR, according to StaxRip terminology). I've read it can be either set in the video stream or in the mux. I don't know where I'm setting it in my own StaxRip or VirtualDub2 encodes. Is it something I should worry about at all? Is there any way to ensure most players get it right? Maybe set both?

    Not much progress done with my filter settings tests. These days have been a bit frantic so I haven't had the chance to do much.

    I'm leaning towards doing my encodes with StaxRip, now that I know how to change the QTGMC settings. It's been useful to learn the basics of AviSynth and even get the filter to work there for on-the-fly testing, and I'll probably keep playing around with it to learn more, but I still have a few problems I'd need to sort out if I wanted to encode everything with VirtualDub2. Unless you tell me I will have the same problem with audio sync even if I encode with StaxRip (it just dawned on me). There's also subtitles, chapters, though I haven't even checked how that works yet.
    Last edited by MGRV; 28th Dec 2018 at 09:21. Reason: I keep calling VirtualDub2 "VideoDub2"
    Quote Quote  
  28. Member
    Join Date
    Dec 2018
    Location
    Spain
    Search Comp PM
    Happy new year, folks!

    As you have helped me so much with this, I figured a little update was due (uninteresting as it may be).

    I did a lot of testing lately. I compared slow vs. fast preset frame by frame, but modifying one setting at a time from the slow preset to match the value of the fast one. The logic behind it was to see if any single setting made a noticeable impact on the way the tooth is smoothed out. It's really hard to notice, as the differences are expected to be minimal, but long story short, it's the combination of settings and not just one variable.

    So, then I checked for any artifacts during playback that could need a slower preset. I can't say I found any. I tried both in my laptop and on a TV. I do see a better quality in slower presets when I compare frame-by-frame, but I have not been able to spot any major difference during playback.

    I also googled to see if there were any other 'practical' comparisons around between QTGMC's different 'speeds', anything that could help me educate myself (or my eyes) further, but couldn't find anything.

    So I guess a few more tests are on the way before I make the call. So far, frame-by-frame comparison sways me to slower presets; playback leaves me undecided; the 'tooth factor' (which I took as an indication of possible detail loss elsewhere) makes a cause for the fast preset. I know this is not only subjective but also very much dependent on the source. But would you advice at all against the fast preset, in principle?

    I'm pretty happy with the Film tuning, by the way. I had never tried it, but I think I'll be sticking with it.
    Quote Quote  
  29. Member
    Join Date
    Dec 2018
    Location
    Spain
    Search Comp PM
    In case anyone cares to read this far at some point, I should say that further testing proved noticeable differences between slower and faster QTGMC presets. Particularly in one of the DVDs, there is some heavy aliasing that benefits from slower settings. I wasn't able to spot anything like it in the first DVD I tried.
    Quote Quote  



Similar Threads