Help with video to create: no-recode compatible, with existing AVC H.264?

[bulgom]

.
Hi,

about ".mkv" intermediate files, you thought you knew what a lamer is? Well... you did not! * But now, you do. I just couldn't remember "MP4Box" extract & mux commands...

So, thanks for that reminder.

After that, being in a state of deep psychodrama, I could no more avoid editing post #19. So it's done, + attachment replaced along.

* Or may be you knew, after all! Because... when I read your "path(s)" advice... ...

OK, I don't know much (at all!), but at least, I'm familiar with paths... [ And sure understand that, by reading what I post, you may wonder to what point you need to explain. ]

My L-SMASH problem is not due to a path error (too bad, btw). I guess I'll have to test that step in another PC... Thank you also for the links & instructions — but I still get the most stubborn and strange reactions, from that plugin: all worst possible error messages:

—— although it worked: once, and once ONLY! ——

Avisynth open failure: Script error: there is no function named "LSMASHVideoSource"
+ GetLogicalProcessorInformation entry point not found KERNEL32.dll. +
VirtualDub Error: Unable to open file: C:\__2\zz_Example__l-smash.avs +
This app couln't start: MSVCR120.dll not found...

Why, did it work, ONCE (earlier today)? Good question...


I could then observe no garbled image, no problem, on yesterday's joined test. So, as you said, my display problem must be related to "FFMS2" itself. [ Besides the mismatch you explained. ]

And now, NO way to make it work again. I "CCleaned", restarted, uninstalled, deleted all registry keys manually, downloaded a new "VirtualDub", reinstalled "AviSynth 2.58", then "2.60". Total mess & no better...

Well, that was..."the~report"!

_________________________



One more "little" problem, I began worrying about:

Desynch. due to "VFR" mode sure gave me hell, 3 - 4 years ago, while helping friends join their cellphone clips — until I decided to recode almost systematically, in that case **.

I just read https://forum.videohelp.com/threads/375244-adding-a-title-to-MP4-clips?p=2418619&viewfu...=1#post2418619


Now:
so far, I only "advertised" the magic "SPS-ID" method to two friend couples, but something tells me I'll have to lecture them soon, on "VFR vs. CFR" ! — because: of course, they ARE going to join VFR (darn) cellphone videos! (otherwise Earth would be Heaven)


But me, I DO-NOT want to have to resynch. their XXX hours of video (knowing that it's often impossible)! Please: anything but THAT!


** At the time, I found "AviDemux" "RESAMPLE FPS" filter (NOT its "Change FPS" filter, of course) to be able to avoid all desynch., as long as I'd modify the FPS value;

AND as long as: I'd also recode AND reasample + reset to "stereo", the audio stream.

plus, while at it: FPS modified to the closest standard & round value (except NTSC FPS), e.g. 29.97 when 28.312 found or other weirdness, or 25.00 instead of 26.413, 15.00 instead or 14.849, etc.

I fact, I know that setting a "std + round" FPS value is not the reason (the fix). I'm quite sure it could be anything, weirdo-est included — as long as it's set to: CFR, of course.

Not only there is no "CRF / VFR setting" in "AviDemux", but it doesn't default to CFR, when recoding AVC videos

— in fact, it DOES: only when all (or part of, but which exactly?..) those heavy duty settings are applied.

I still don't see any reason why it would fix a desynch. (not within ONE video only, of course = different case), that occurs when joining several VFR bitches, just because I set it as above. But again... the fact is: it does (not "VirtualDub", "for instance"). "MediaInfo" then displays: Frame rate: constant.

[poisondeathray]

It's a big headache.

They type of VFR commonly used in consumer devices is done by using timecodes. There are actual dropped frames (it goes lower fps in sections, higher in others), but the timecodes keep everything in sync. The actual number of frames is lower than if it was shot with CFR in the first place (ie. there are missing frames). For minimally variable videos, you can just treat it as one CFR, and it will be very slightly in and out of sync in sections but for the most part still be ok. The problem is with those videos that have large deviations, or many of them.

VFR doesn't affect the actual joining of elementary bitstreams - they don't really have a "framerate" per se. But to keep everything in sync, you would have to shift the timecodes. This is not fun to do, especially if you have very variable samples (ones that fluctuate a lot) because you would have to offset and shift everything. To understand this more, have a look at some extracted timecodes - for a VFR video it will have an entry for every point the frame rate changes.

When you re-encode and resample the FPS, what you are doing in most programs is inserting duplicate frames in approximately the correct places in time - you're converting it to CFR to that specified frame rate (almost as if the FPS wasn't lower in those sections in the first place ie. as if it was CFR to begin with). For very variable videos, it will audibly go in/out of sync slightly in sections

Even completely re-encoding can sometimes be a problem, depending on how you do it and what software you use. VFR can cause many problems in editors for example

[bulgom]

.

I sure hope one those cellphones companies to stumble on this post, someday soon (but I know they're aware... I must be the XXX,XXXth user to complain).

Yes, as you explained before; I had read that and understood OK.

Except: I don't know how to extract timecodes.

It's a big headache.

for a VFR video it will have an entry for EVERY point the frame rate changes.

When you re-encode and resample the FPS, what you are doing in most programs is inserting duplicate frames in approximately the correct places in time

Although "AviDemux", under certain settings, seems (?) to take care of that, I wonder if exists some utility that would (specifically) solve THAT problem, precisely by duplicating frames — as, I guess, "an entry for every point" results in a huge / more than tedious list of digits...

_____________

What I've also read about "VFR" (dunno if only reason) is: cellphone builders wanting to save space + connections bandwith, consider that a scene that's very still can well framerate at, say, 10 FPS or even less...

Good idea... on the paper! In pratice, I sure experienced the "side" effects!

"Another" absurdity: since cellphone companies now provide people with HD, Full HD & even UHD, one of the 1st things that come to consumers' mind is (precisely): JOINING those clips (since their resolution at least, "feels" professional, wanting a "long footage" comes in naturally)!..

PLUS one more, that I can't get over: HD + Full HD + UHD on one "side"... AND 64 kbs monophonic audio on the other "side"! Few years ago, I was very disappointed to observe that, especially on expensive gadgets such as "iPhones", and even on "Canon" movie option of still cameras.

Knowing cellphones are equipped with TWO lenses + resolutions (front & rear), how come in most cases, not two microphones?, and of course, at least one 160 to 256 kbs audio optional setting...

[poisondeathray]

Except: I don't know how to extract timecodes.

It's explained everywhere, just use search ... You can use ffms2, or mp4fpsmod, or mkvextract, or ffmpeg, or flvextract, etc.. probably a dozen different ways



If you are unfamiliar with timecodes, to create VFR , you mux with timecodes. e.g. mkvmerge for mkv or mp4fpsmod for mp4 etc...

Read the mkvmerge documentation about the different types of timecodes eg. V1 vs. V2 vs. V3 etc... It explains quite a bit there. V1 is easier to read for humans, because it uses ranges