Framerate and synchronization issue (29.97FPS)

[abolibibelot]

Hi everyone,

I almost finished editing a video which is 1h19m30s long. It mixes footage in 29.97FPS (first part) and 25FPS (second part). I'm using Magix Video Deluxe (abbreviated as “MVD” below).
At first I set the project in 25FPS, as it's a more common framerate here, and it seems less tricky to deal with an integer value, rather than a decimal one (apparently I was right on that part !). But I had a problem : the 29.97FPS camera had a defect (stabilizer not working properly) so the footage has random instances of vertical jerkiness (not correlated to the normal shakiness when it was shot handheld -- or it would mean that I have “vertical Parkinson disease”), and I had to make intensive use of the “stabilizing” function available in MVD. But it wasn't working properly, I couldn't understand why (I had already used this function which normally works very well), until I realized much later (I had abandoned this project for months in the mean time, mostly because of this issue) that the stabilizing function was correlating the corrective parameters with the output framerate (25FPS) instead of the input framerate (29.97FPS), and was messed up because of the 25FPS setting. When I changed the parameter of the project to 29.97FPS the stabilization worked nicely.

Then, when I made the final export (in AVI MagicYUV, 29.97FPS), the resulting video file apparently had a wrong index, according to VLC Media Player, and was definitely out of sync, no matter which player I tried. If I let VLC correct the index it would play fine. I tried loading it in VirtualDubMod (where it also played progressively out of sync) and re-exporting it (Direct Stream Copy) : no “wrong index” message this time but it was still out of sync. I thought maybe it was because the project was first set to 25FPS then changed to 29.97FPS (and I dreaded having to start it all over again in 29.97FPS if that proved to solve the problem), so I created a new project in 29.97FPS and made a series of tests : a 20sec. export would always play directly (too short to verify the A/V synchronization), but a 30sec. export would always have this “wrong index” error in VLC. I tried exporting in 25FPS (from the original project) : it would play fine (no “wrong index” error, no sync issue), but then the stabilized segments would be jerky (same if exported in 30FPS), for the reason explained in the first paragraph. I also tried copying the source files in another folder and letting MVD re-import them in a new 29.97FPS project (thus re-creating the frame table or whatever is stored in the .m2ts.H0 files) -- same results. It seems like even in a brand new project set to 29.97 with 29.97 source files, MVD insists on converting the framerate somehow, adding duplicate frames at a certain interval (which I haven't yet identified).
The whole movie should be exactly 1h19m30s, which is 4770 x 29.97 = 142956.9 frames, I guess it should be rounded to 142957, but it ends up being 1h19m31s80 or 143011 frames (it's not correct for 30FPS either : that would be 143100 frames). (I couldn't find a way to set Magix Video Deluxe to display the number of frames instead of the duration.) If I export 30s and open the file in VirtualDubMod the video appears to be 901 frames long instead of 899 (as for the duration, the video is 30.06s, the audio 30.05s). I tried exporting the begining segment (11m03s) in video only and audio only (WAV), then mixing the two with AudioDub function in Avisynth : it's still progressively out of sync. I examined a segment with constant motion in VirtualDubMod and noticed a duplicate frame (17304 = 17305) but couldn't find the next one (the idea was to try to decimate one frame every X frames and see if that would correct the synchronization issue) ; my guess is that it would be between 20s and 30s later (which would explain the “wrong index” warning), but that would be tremendously painstaking to do manually.

So, questions :
- Is there an explanation to this weird behaviour ? Is it common to other NLE softwares when dealing with 29.97 video, or is it a bug specific to MVD ?
- How can I export the whole thing with a correct duration and perfect synchronization, in order to compress it (using MeGUI) and be done with it ? If there's indeed an issue with MVD which prevents it from exporting to 29.97FPS with a flawless synchronisation, maybe an Avisynth function would do the trick, like automatically detecting the duplicate frames and removing them ? What would be the best method to do that ? (I haven't tried exporting directly in MP4 within MVD to see if the output would be correctly synchronized, but I would definitely prefer to do it with the intended workflow so as to benefit from the best possible quality / size ratio offered by x264 / qaac.)

Thanks to those who will help me solve this thing.

[pandy]

In real life valid refresh rate is expressed as 30000/1001 not 29.97... repeat all your steps with correct framerate.

[abolibibelot]

Well... thanks for this first reply, but Magix Video Deluxe doesn't provide a “30000/1001” export setting, and the framerate setting doesn't allow more than two decimals (so I can't set 29.97002997 either). So only 29.97 is available. But still, the difference should be negligible, shouldn't it ? (29.97 is 0.999999 of 30000/1001, so for 4770 seconds the discrepancy should be less than 1/100s if my calculation is correct, hardly enough to produce a noticeable desynchronization.)
Anyway it would seem like the duplicated frames are present within the timeline already, so something is wrong in the way the 29.97FPS source files are imported and dealt with in the first place.

What else could I try ? Is my idea to detect and remove duplicate frames with Avisynth a good one, and how could I proceed – with which plugin – to do that automatically ?

[manono]

Is my idea to detect and remove duplicate frames with Avisynth a good one...

Possibly. How about a short sample (10 seconds or so) from the source so we can have a look?

[abolibibelot]

Well, I'm about to sleep now ! ;^p I'll do that in the afternoon.
By source you mean the file which has been exported by MVD, right ? Or the actual source files before they were edited ? (Those are .m2ts from a Panasonic TZ10/ZS7 compact camera.)
Would 10 seconds be enough to detect what's wrong ? Should I let that sample in lossless AVI, or is it fine if I recompress it for a quicker upload ?

[pandy]

Well... thanks for this first reply, but Magix Video Deluxe doesn't provide a “30000/1001” export setting, and the framerate setting doesn't allow more than two decimals (so I can't set 29.97002997 either). So only 29.97 is available. But still, the difference should be negligible, shouldn't it ? (29.97 is 0.999999 of 30000/1001, so for 4770 seconds the discrepancy should be less than 1/100s if my calculation is correct, hardly enough to produce a noticeable desynchronization.)

Calculation are correct but behaviour for player can be variable and you didn't put this in a equation... IMHO .

But OK - this is really curious case - and here manono is expert.

[manono]

By source you mean the file which has been exported by MVD, right ?

I suppose, if it's in some sort of standard video format. Being able to compare it with the real source might be useful as well. But you understand, don't you, that just removing duplicate frames also adjusts the framerate with the result that the video will be the same length as originally and won't fix your synch issue? You can, of course, adjust the framerate if you like.

For example, a standard IVTC on a 29.97fps source removes one frame in every five, results in a 23.976 framerate the same length as the original 29.97fps one.

Would 10 seconds be enough to detect what's wrong ?

If the cycle is reasonably short and it repeats regularly, yes.

Should I let that sample in lossless AVI, or is it fine if I recompress it for a quicker upload ?

It's progressive already, right? So recompressing it should be fine.

[abolibibelot]

I suppose, if it's in some sort of standard video format. Being able to compare it with the real source might be useful as well. But you understand, don't you, that just removing duplicate frames also adjusts the framerate with the result that the video will be the same length as originally and won't fix your synch issue? You can, of course, adjust the framerate if you like.
For example, a standard IVTC on a 29.97fps source removes one frame in every five, results in a 23.976 framerate the same length as the original 29.97fps one.

As you can see in the screencap I showed above (VirtualDubMod informations for the exported AVI file) there's a significant discrepancy between the duration of the video stream (1:19:31:80) and the audio stream (1:19:30:01 as it should be, i.e., as it appeared in the timeline -- well, it was exactly 1:19:30:00, so it's just 1/100s more which is negligible). So I figured that if I could remove extra duplicate frames while still keeping the framerate at 29.97 (or 30000/1001, whatever), this could correct the sync issue without affecting anything else.
(Of course it plays fine with no sync issue in MVD's timeline. I'm still trying to figure out if it could be related to internal settings in the software, or the way the source files are imported / treated within it.)

Source file (M2TS recompressed to Xvid), frame 175 to 1175 (opened in VirtualDubMod with an AVS script using FFVideoSource/FFAudioSource) :
https://www.mediafire.com/?2q28al9yyck57qf
Exported file (lossless AVI recompressed to Xvid), frame 17300 to 18300 (the picture is slightly cropped because of the “stabilize” function) :
https://www.mediafire.com/?2q28al9yyck57qf
I selected a 1000 frames segment which is about 33 seconds. It begins at frame 17300 of the exported video file (~9m30s), so the duplicated frames I noticed earlier are the 5th and 6th (exported file). Since there are 54 extra frames for a total of 143011, there should be one extra frame every 2648 or so, right ? Would it be enough to cause a noticeable desynchronization, or is it something else ? At this rate, at about 9m30s (from the begining of the edited movie) there should be already 6 extra frames, which amounts to a 200ms shift, which seems consistent with the shift I can notice in this sample (see/hear the « Ah ! » at the begining for instance, the audio is slightly ahead of the picture). If that's indeed the case, the sample provided here would be too short to search for the next or previous extra frame (but in this case, at +2648 there's a montage of still pictures, and at -2648 there's a short section which is slowed down to about 50% so every frame is duplicated, and before that it's a part with very little motion so it's tricky to notice a repeated frame visually).

[EDIT] Frames 25291 & 25292 are duplicates. Frames 27953 & 27954 seem to be duplicates too. Then 30615 & 30616, then 33277 & 33278, 35940 & 35941, 38602 & 38603... It would seem that the periodicity is approximately one repeated frame every 2662 (unless there are others in between but it would be unlikely). Does this correspond to a known particular ratio / shift / whatever value ?
[EDIT] The first repeated frames (found by counting backward, removing 2662 each time) seem to be 1331 & 1332 (so the pattern doesn't begin right at the begining of the file).


I found some discussions regarding such issues with Magix editors and 29.97FPS but nothing conclusive. This one seems to present a similar issue (the original poster likewise set a project in 25FPS at first and then changed it to 29.97) -- but it's in German, and I don't read German fluently (which really is a shame as that was the first foreign language I studied in school, before english...), an Adobe editing software was used, and it doesn't mention a synchronization issue (only jerkiness).
https://www.slashcam.de/info/Material-25-fps-Film-29-97-Problem---125626.html
http://www.slashcam.com/EN/info/25-fps-film-material-issue-29-97--125626.html (automatic translation in english)
In this other discussion (in very bad french) the OP complains that MVD v18 (a more recent version, mine is v17) exporting to 29.97FPS instead results in a 25FPS video with a huge synchronization issue :
http://tinyurl.com/bugMVD18

[manono]

File Blocked for Violation.

The file you requested has been blocked for a violation of our Terms of Service.

Still have questions, or think we've made a mistake? Please contact support for further assistance.

It says that for both.

[abolibibelot]

I don't understand, both links work fine for me... (Firefox, AdBock and NoScript activated)
But I did make a mistake and posted the same link twice. So here are the links as they should have been :
https://www.mediafire.com/?2q28al9yyck57qf
https://www.mediafire.com/?g13jn7lm81m59p2
Here are the fully developped links :
Source file (M2TS recompressed to Xvid), frame 175 to 1175 (opened in VirtualDubMod with an AVS script using FFVideoSource/FFAudioSource)
Exported file (lossless AVI recompressed to Xvid), frame 17300 to 18300 (the picture is slightly cropped because of the “stabilize” function)
And here are two alternative links if the Mediafire ones still fail :
http://www.cjoint.com/c/FHzdY022vPy
http://www.cjoint.com/c/FHzd1ivQGDy


Anyway, from the duplicate frames I already identified (apparently one every 2662-2663, starting with 1331 & 1332), would you have some clue about what's going on and how to fix the issue with as little hassle as possible ?
The tricky part may be the segments edited from 25FPS source files (about 35 minutes at the end) which, if I understand this correctly, will contain legitimate duplicate frames in order to convert the original framerate to 29.97FPS.

[abolibibelot]

Further tests :

– If I create a brand new project in Magix Video Deluxe v.17, set it to 29.97FPS, add one 29.97FPS M2TS source file which is 18m50s long, and export exactly the first 10m00s00 (according to the duration displayed in the timeline), the resulting file has exactly the same issues as before : wrong index according to VLC Media Player, duplicated frames every 2662-2663 frames, the first one being frame #1332, and progressive video/audio desynchronization. The duplicated frames apparently do not appear in the timeline, they're added only during the export. (Same result with MagicYUV, UtVideo, Lagarith.)

– If I re-import that exported file into the timeline, it is displayed as having a duration of exactly 10m00s00, right at the end of the selection previously made (when zoomed to the maximum, the grid unit being one frame, there appears to be a “partial frame” sticking out after the selection mark, if that makes sense). But the two files are not perfectly matched : to match the picture I have to shift the source video by two frames to the right at the begining (then the audio doesn't match). Is it to be expected when losslessly recompressing from a compressed format which is not “key-frames only” like M2TS ? Then if I go at the first duplicated frame, around 0m44s, I have to shift the source by one more frame to have the picture from both videos matched. Then at the end of the 10m00s export the total shift between the two videos to have the picture matched is 9 frames.

– If I change the program parameters from PAL to NTSC, it doesn't change anything.

– If I select “Automatic creation of a frametable during import” and re-create a project and re-import the same file and re-export 10m00s00, same result.

– If I export to uncompressed AVI : I no longer get the warning in VLC, but there are still repeated frames at the same spots. But strangely, the picture is matched with that of the source footage at the begining (I don't have to shift the source by 2 frames), as opposed to the losslessly compressed AVI exports. Once I have shifted the uncompressed AVI by two frames relative to the losslessly compressed AVI, both are matched right through the end (so by the end the shift is 7 frames between source and uncompressed AVI, instead of 9 for losslessly compressed AVI).

– If I let VLC rebuild the index (which is only temporary) before reading the losslessly compressed export it then displays a duration of 2m23s (instead of 10m00s), so it's too short to verify the synchronization (which is progressive thus barely noticeable during the first few minutes) ; it otherwise plays normally until right before the end (at 2m23) where it suddenly cuts to a very short segment which is near the end of the 10min, and stops.

– If I export in MP4 : like the losslessly compressed AVI, there's a shift of 2 frames at the begining, the duplicated frames are at the same spots, but there are 7 extra frames at the end (if I re-import the MP4 in the timeline the duration is displayed as 10m00s07f), just as many as the number of duplicate frames (which is deduced from the shift relative to the source to get the picture matched).

I don't know yet what I can make of all this weirdness... or, again, how I can fix it...

Below are two screencaps from MVD corresponding to the matching test above (tracks 1-2 = source M2TS video / tracks 3-4 = losslessly compressed AVI 10m00s00 export / tracks 5-6 = uncompressed AVI 10m00s00 export ; the tracks are shifted relative to each other so that the picture matches for all three near the 10m00s00 mark ; on the first image the grid unit is 5 frames, on the second image the grid unit is 1 frame).

[manono]

The first link is still blocked. I got the second of 1000 frames but don't feel like working my way through 1000 frames looking for duplicates, for whatever reason. I don't use Magix Video Deluxe. Perhaps someone else can help. Good luck.

[abolibibelot]

The first link is still blocked. I got the second of 1000 frames but don't feel like working my way through 1000 frames looking for duplicates, for whatever reason. I don't use Magix Video Deluxe. Perhaps someone else can help. Good luck.

Did you read my latest posts ? I already identified a pattern of 1 repeated / duplicated frame every 2662 (or 2663), starting with frame 1332. The last post with various tests and screencaps from the timeline should make that clear enough. I've done most of the guess work on my own. Now I'm just looking for a way to fix the issue (i.e. accurately remove those extra frames so as to hopefully restore the synchronization), using Avisynth or something else (and if possible, understand what's going on, but that's not the most important right now).
Part of what makes this tricky is that the periodicity appears to vary very slightly (2662-2663 as far as I could see so far), so a simple “decimate 1 frame every X frames” function wouldn't be accurate (it would leave duplicated frames and remove non-duplicated ones). The other tricky aspect is that there's 25FPS footage (and a few slowed down sections) which will legitimately contain duplicate frames once exported as 29.97FPS (I'll make some tests to analyze the pattern of the framerate conversion), so a method that would automatically scan the whole video for duplicated frames and remove them all would result in a severe desynchronization in that part. So how can I proceed, short of locating and removing manually all 54 extra frames ?

[manono]

If the original video doesn't have the dupe frames and only the output does, it seems to me the answer is obvious - use a different editor.

Yes, the duplicate frames can be removed. There are several ways. One might be MultiDecimate. Others might include Dedup and ExactDedup.

[abolibibelot]

In this case, the movie in question is virtually finished (perhaps a few things to polish but the bulk of the work is done), I can't start all over again now with a brand new editor, which I won't master at the same level before a few weeks or months, and which will no doubt present a whole new bunch of issues for any reasonably complex project (and this one includes quite a few complexities – for instance I need a good stabilization function, I guess not every editor features one). I'm thinking about trying a more recent version of a Magix editor, which may or may not solve that particular issue (hoping that it will open the current project file without messing anything in the process, and that it won't have other bugs or missing functionalities compared to the version I'm used to).

Again, the main question is : is there a way to salvage this thing as it is, and get something that is watchable ?
And so, you mention a few possible methods, good, but which one would work best in this particular case, and what would be the most efficient way to achieve what I want ? I already mentioned what makes this task not-so-trivial :
1 - the periodicity of the duplicated frames is not perfectly regular ;
2 - there are parts with legitimately duplicated frames that should not be removed.
So, how could I deal with this with one script ?

[manono]

1 - the periodicity of the duplicated frames is not perfectly regular ;

It doesn't matter. They are two-step processes, an analysis pass followed by the dupe removal using parameters you set

2 - there are parts with legitimately duplicated frames that should not be removed.

Even dupe frames in the original video should have slight differences while dupe frames created by the editor should be true dupes. I used to use MultiDecimate with some regularity and haven't used the others. For MultiDecimate you'd go over the file created after the first pass comparing a few of the dupe frame metrics to see if there are differences between the original ones and the added ones. Never know till you try.

[abolibibelot]

The thing is, I created a new thread here because it came to a point where I felt like I couldn't solve this on my own, or I would spend a lot of time (and I've already wasted way too much time on this) trying different methods with a bunch of tools I never used, having to first learn how to properly use each one of these, then running tests, and other tests, discovering in the process unexpected issues that perhaps regular users of said tool are well aware of and know how to circumvent... I was hoping that I could learn something new and readily efficient, relevant to my particular problem, some practical, informed knowledge from people who know their stuff, in a word, a little help (that's half of the forum's name isn't it ?), instead of wandering like a poor hungry dog from one barely relevant Google hit to the next with no meaty bone in sight any time soon. Sorry but I feel a bit frustrated here. I did most if not all of the tedious work on my own, I don't expect to be spoon-fed the complete solution, but I would dearly appreciate a specific hint toward it, with a tip or two to boot. Someone said you were “the expert” for that sort of things, a reputation you must have earned somehow, so could you please give me a specific example of how you would use MultiDecimate – since that's the one tool you usually use – in a case like this, and maybe what can possibly go wrong, and how to deal with it ?

As I've already said : about 35 minutes at the end of this movie are edited from 25FPS source footage, so I guess that to convert this to 29.97FPS the editor will have to generate exact duplicates of some frames, and I sure don't want to remove those. I think it's pretty smart on my part to foresee such an issue, and ask for a possible workaround beforehand, instead of trying and failing and asking why at every single step. That kind of effort is rather rare on support forums like this and could/should be appreciated.

Thanks.

[manono]

At the risk of sounding rude, let me get this straight. You decided to use an inferior editor (which you might not have known to begin with was inferior) to create a project that ended up a disaster. You don't want to begin all over again and do it right because it'll take too much time, yet you spend hour after countless hour counting frames, searching the internet, trying this and that to try and undo the damage already done, some of it by you.

Just stretch the out-of-synch audio to the exported 29.97fps length and be done with it. Any WAV Editor (including the freeware Audacity) can do that. The repeated frames are at regular intervals so it should be back in an almost perfect synch if done correctly.

I don't know if you ever mentioned the final format, but if for DVD you certainly don't want to duplicate frames to bring the 25fps up to 29.97fps. Keep the 25fps portions separate and encode as 25fps with 3:2:3:2:2 pulldown afterwards to output 29.97fps. You'd do that using DGPulldown and then you'd rejoin the two sections during authoring.

[abolibibelot]

At the risk of sounding rude, let me get this straight. You decided to use an inferior editor (which you might not have known to begin with was inferior) to create a project that ended up a disaster. You don't want to begin all over again and do it right because it'll take too much time, yet you spend hour after countless hour counting frames, searching the internet, trying this and that to try and undo the damage already done, some of it by you.

Is it common knowledge that Magix Video Deluxe is an “inferior editor” ? When I first looked for one, 4-5 years ago, after reading a few reviews, it seemed like the best choice overall for me, as a beginner aspiring to quickly learn more than just the basic stuff ; it was generally described as having a good balance between ease of use and richness in features, with a good stability, a broad compatibility with available formats -- all of which proved to be true (for instance it could import 1280x720 AVCHD and play/render them smoothly on a 2009 PC based on a dual-core Intel CPU with motherboard integrated GPU and 4GB of total RAM, whereas other “big names” of NLE softwares were apparently struggling at the time even on beefier machines), and I've been mostly satisfied with it. No program of this complexity can be perfect, and even the ones aimed at professionals have their drawbacks. I'm willing to try something else, but I've read about bugs or weird behaviours or important features lacking concerning pretty much all the most prominent softwares in the field.
Honestly, if you discovered such an issue at the very end of a project, when everything has been finely tuned, every defect treated to the best of your knowledge with the available tools of an editor you know well, would you start all over again with a new one in the (very unreasonable) hope that it fixes that one issue and does everything else just as well with no hassle ?

The project did not “end up a disaster”, and I don't feel like I “damaged” anything, there's just a weird bug when exporting to 29.97FPS framerate (it's a german company, I guess their products are mostly aimed for PAL standard, it could explain that 29.97FPS export was treated as an afterthought). I really don't understand the need to be so rude just for the sake of justifying your unwillingness to actually be helpful. Your first reply could have been : “Yes it could be done with MultiDecimate, with such a script [EXAMPLE SCRIPT], just beware of this and that, let us know if it worked...” That would have been helpful (to me and possibly other people having a similar issue later on). I could have tried something worthwhile right away, and maybe have this solved and finished by now.

Just stretch the out-of-synch audio to the exported 29.97fps length and be done with it. Any WAV Editor (including the freeware Audacity) can do that. The repeated frames are at regular intervals so it should be back in an almost perfect synch if done correctly.

That seems like a very “quick and dirty” way of fixing this. At worst, now that I found their interval pattern, I could remove those ~50 extra frames by naming each one of them in a decimate script (it's quite a lot but still manageable), and keep the audio (which is fine as it is) intact. What I was looking for here was an easier and more elegant way of achieving this.
But -- see below -- updating MVD seems to have fixed that issue... while bringing a new one...

I don't know if you ever mentioned the final format, but if for DVD you certainly don't want to duplicate frames to bring the 25fps up to 29.97fps. Keep the 25fps portions separate and encode as 25fps with 3:2:3:2:2 pulldown afterwards to output 29.97fps. You'd do that using DGPulldown and then you'd rejoin the two sections during authoring.

I already said what the final format will be : x264/aac in MP4 or MKV container. It doesn't make sense to encode this as interlaced (if I correctly understand what you mean and what 3:2:3:2:2 actually does).
In this case, since the 29.97FPS part and the 25FPS part are separate (i.e. not mixed together, two distinct parts of the movie), would it be wiser to treat them as two separate projects, keep both of them at their native framerate, encode them separately and then stitch them together ? Is it compliant with standard specifications to have a sudden framerate change in the middle of a video file ? Or what would be the best way to treat this, so as to have the smoothest playback for all portions ? (I have not yet tried to analyze how MVD converts 25FPS to 29.97FPS.)


***************************

Now, I installed Magix Video Deluxe 2016 on a Windows 7 system (the former version is still on my XP partition, I prefered to keep it there instead of updating it in-place, in case that move would hurt more than heal), and I re-made the test of exporting a 10:00:00 section of a 29.97FPS M2TS to lossless AVI : there's still this VLC Media Player warning about an erroneous index, *BUT* the resulting file seems perfectly synchronized this time, and according to VirtualDub the video and audio streams have exactly the same duration (10:00:03). The project file from MVD 17 is correctly recognized, there are a few things that appear a bit messed up (still pictures with the wrong aspect ratio for instance), but it shouldn't be too long to fix.

Yet, there is a new issue regarding file import this time : it can't import raw YV12. I used Avisynth Virtual File System to load some of the 25FPS files which badly needed pre-filtering (filmed in an apartment in front of a window resulting in a very dark picture generally, with many changes in exposition -- I used Autolevels and HDRAGC, which already improved it significantly, but of course it's not perfect and I'd be interested to know of other tools suited for that kind of correction). It worked perfectly in MVD 17 on Win XP, now what is missing here ? I installed ffdshow and tried to tweak some settings, to no avail. I tried disabling internal import filters in MVD 2016, doesn't work either. The virtual AVI files are correctly opened by VirtualDub, and even Windows Media Player. It works if I add "ConvertToRGB()" to the script (which more than doubles the size of the virtual file, and may or may not make it heavier to process).

[manono]

Is it common knowledge that Magix Video Deluxe is an “inferior editor” ?

I didn't say it was common knowledge. I know nothing about it. You did say you found some references about similar problems when searching. You were blaming it for adding the duplicate frames and messing up the audio synch.

I gave you three different ways to remove the duplicate frames yet you seem unwilling to try. MultiDecimate is not "the one tool you (I) usually use". I haven't used it for probably a decade - since I was pulling dupe frames out of silent films. Relearning it would probably entail nearly as much work as you learning it for the first time. Its use is detailed clearly at the link I provided. You can lead a horse to water...

I already said what the final format will be : x264/aac in MP4 or MKV container.

Okay, sorry. You're kind of ... wordy, so forgive me if I forgot or it never registered to begin with. Depending on how much movement is in the video, I'd probably interpolate the new frames between 25fps and 29.97fps, rather than just adding duplicate frames. If it's a complex video (lots of movement), then ChangeFPS might be the way to do it, as long as you understand it will play pretty jerky. There are many interpolation scripts. Here's one:

super=SVSuper("{gpu:0}")
vectors=SVAnalyse(super, "{}")
SVSmoothFps(super, vectors, "{rate:{num:1200, den:1001}}", url="www.svp-team.com", mt=1)

It requires both SVPFlow1.dll and SVPFlow2.dll.

[abolibibelot]

Relearning it would probably entail nearly as much work as you learning it for the first time. Its use is detailed clearly at the link I provided. You can lead a horse to water...

Well, again, someone called you “Teh Expert”, sorry if, trusting that statement, I expected you to know most if not all of the various Avisynth plugins related to framerate issues / duplicated frames removal, and which one would be the most relevant for a case like this. (I don't blame you for not knowing more than you do, but then it's kinda frustrating if someone who barely knows one such filter, admitting that I'll learn just as much by RTFM, is considered an authority on this subject on this forum. It's akin to being told about an expert in boxing, and then meeting a guy who reluctantly mutters that he used to be involved in bar brawls and stuff but that was long ago, and gives me the title of a good book where I can learn everything about the uppercut.)

Okay, sorry. You're kind of ... wordy, so forgive me if I forgot or it never registered to begin with.

Well, I tried to be thorough so as to avoid any misunderstanding (as there seems to be quite often wrong assumptions, based on personal habits or prejudices, about what a person is trying to achieve, judging from past experiences), especially considering that english isn't my first language and I may not always use a perfectly accurate phrasing, therefore I prefer to be extra careful so I don't have to explain myself twice or more ; but indeed, if this results in (presumably knowledgeable) members not reading the details cuz they get bored after five lines, it defeats the purpose... I'll try to formulate my questions in haikus next time. Anyway there's a search function in any decent browser, on this page it gives hits for "MP4", "x264", but, until you mentioned it, none for "MPEG" or "DVD". And even at my rather modest level, when I read a request from someone else, if I'm willing to try and help, I'll first make an effort to have a broad picture of what needs to be achieved, and at what knowledge level that person operates, so as to avoid giving too obvious and/or misleading answers (which is all I got here, hence the aforementioned frustration).

In this case, the screencap I provided in my second post should have been enough (without even reading my explanations) to have a good hint at what was wrong and what to do next : there was a significant discrepancy between the total number of frames and what it should have been considering the duration and the framerate. And with a simple calculation (which I later made on my own) it should have been obvious that providing a 10s sample wouldn't provide any new insight (the interval would be higher), yet you asked me to provide just that, and then admitted being reluctant to do any serious analysis of it (I thought that you would know of an efficient method to automatically detect duplicated frames and their periodicity and whatever other useful information, finely tuned and ready to run, otherwise I wouldn't have gone through that hassle and waited for an answer). At least I did do active research on my own, and writing about the issue may have helped me to sort it out, but I barely got any new input. Maybe I spent “hours after countless hours” trying to solve this, but that's why I had this foolish idea to request some insight and support from someone with more experience, who would see through this in a second rather than painstakingly elucidating the crux of the matter.

Depending on how much movement is in the video, I'd probably interpolate the new frames between 25fps and 29.97fps, rather than just adding duplicate frames. If it's a complex video (lots of movement), then ChangeFPS might be the way to do it, understand it will play pretty jerky.

Shouldn't an editor ideally deal with framerate conversions in a way that produces a watchable result ? (I know I'm already in a case that's not-so-ideal but still...) Apparently MVD did indeed repeat about 1 frame every 5 frames for that 25FPS part (with probably a different pattern once in a while to -- theoretically -- obtain 29.97FPS, and probably non legitimate extra frames here too that account for the global discrepancy).
I can't use Avisynth to convert the framerate before importing (with the sudden changes in exposure (1) I had to mix the source footage imported directly and corrected internally with basic gamma/contrast adjustement, with the source footage pre-corrected using Avisynth and loaded through Avisynth Virtual File System ; there are already 4 such virtual files that must be loaded when I open this project and it's pushing my system close to the limit), so I'd have to add this at the encoding stage, hoping the end result would be worth it. Besides, wouldn't it affect the picture quality, and significantly increase the encoding time, to interpolate all the extra frames ?


(1) I know, I did some “damage” here, but I didn't know it at the time, on Christmas day of 2013, when casually filming my grandmother talking with my mother and an aunt (essentially to have something to show to my disabled brother who hadn't seen any of them for more than ten years, and then I only sent him a 1m30s excerpt where “Mémé” said hello), that I would put that footage at the end of a movie about her funerals which happened to happen exactly two years later, on Christmas day of 2015 (I couldn't get my brother, who's like a child in his mind and was deeply affected, to come with me and be there, so once again, I promised him I would at least make a video, “to remember and to let go”, as Christian Shephard said at the end of Lost...).



**************************


So, now that I've found a possible solution which no longer requires a tricky decimation process, would anyone have a clue about this YV12 problem ? Or do I have to create a new thread for that particular question ?