I think it just goes to show the difference between 1 frame behind and 2 frames behind is more subtle than 1 frame behind and right on the money. But as always, thanks KZfaq.

And for those wondering (like myself) what an "MKBHD" is, he's Marques Brownlee and the video is called "This Is What Happens When You Re-Upload a KZfaq Video 1000 Times!" watch?v=JR4KHfqw-oE

Generally how it goes

Well i hear a horn before i see a car infront

That's also when my friend yells "Yo mama's hand. Drop your butt and run!"

And that's just the PC side. Now watch this video on a set-top box (Apple TV or Amazon Fire or ...). The audio goes to the 7.1 digital amplifier, which sends it to the speakers with very little delay. The video gets send to the TV, which does its picture processing, frame interpolation and and and, so it gets delayed by 120ms before it is shown.

I like your brief insight to the audio domain. Do you think there is a solution for this problem? Also, recently I was listening to the podcast about the switch from analogue to digital sound production, and latency was the main problem in this switch. I believe the author said about live broadcasting that: "we are living the same moment but at different times"

@@antontserepanov9720 Yes there are solutions, as when it's bad it's often due to sloppy programming. But it can't be perfect, say you're playing a videogame through wireless headphones, you wouldn't want the image delayed to match the headphones, it'd make the game even worse. Delaying is only ok for non-interactive stuff. I don't think there's any good reason to go back to analogue for sound production. Digital effects with big latencies exist, but they're rare enough (& don't have any analogue counterpart anyway), otherwise a good pro setup these day, with audio through ASIO drivers, can easily reach

Lip sync is a very big struggle for me when it's off... Trying to get it back is hard nut to crack

From what I've read, it's a problem with encoding h.264 in general in regards multiplexing (which is how audio and video are separately compressed and then synched together). It would take me a lot of research to understand it.

@@FilmmakerIQ Ok thank you. To be sure, I tried to replicate your test here on a 4k H264 main 5.2 59,94 FPS video montage (2 min) which was already uploaded last week. And i am surprised at the conclusion if this is correct: - i compared the video that i had to export in premiere pro 2 times instead of 1 to layer another text effect properely on it - I noticed the audio of the 2x exported video started 2 frames EARLIER (so no delay, but faster) than the audio of the original unexported video when these video's were both exactly lined up (FYI timeline is split up in frames for me, not in ms that you mentioned at the end of this video). - Then I downloaded the youtube video via a website that allowed 1080p downloads (i know somewhere in the comments you said that a random youtube downloader site wasn't totally accurate to check, but i can't remember the right youtube downloader) and this video is near identical to the original. (so it went back 2 frames at this 59,94 FPS video and another really small portion of the frame. You could almost say then that it is identical to the 1 frame of the 24/23,94 FPS you talked about). - This would mean that the youtube delay actually canceled out the early 2 frames of the export - Then I watched it on my phone for the first time on youtube (not youtube studio) from a short distance and everything felt really good how i wanted it to be. If this website downloader & test is 100% accurate, that means if I only exported it once, as usual , KZfaq would have only delayed the audio 1 frame later (because 1 frame of audio was earlier from premiere pro export, and the video from the youtube downloader site delayed the audio 2 frames) with the H264 file (converted from .webm coming from PS5). Then the conclusion would be that 1 frame for a ] 60 fps video is pretty much negligible. So weird, but if this is true, it sounds like every situation can be somewhat different.

Yes, the audio beat felt more laid back and not as sharp.

Feel the same

yea, i get you. its like when you park the car and need to turn down the music so your other senses can kick in. similarly, if the cut is right on the beat, it feels a little chaotic for a milisecond or few, because theres a lot of information at the same time kind of thing.

@@krytxps yeah exactly, the audio and visual information sort of destructively interfere and the overall result is diminished

@@MrSb192 yup, diminished is the word

The biggest mistake people make is comparing the human visual system to cameras. Eyes are nothing like cameras - in fact the cheapest plastic lens you can buy would outresolve the human eyeball by a huge margin. It's the brain that makes up most of what we see... And yet, when I get attacked on this subject and quite rudely, it's almost always in respect to this fundamental mistake.

I don't know if it's sad... if KZfaq is doing that I would rather know and anticipate it than rather not know ;)

@@FilmmakerIQ hopefully, it's doing it consistently! :D

In other words in both cases you preferred the cut one frame ahead of the beat ;)

That's a different issue regarding things like attack and resonance as well as auditory dynamics. Screaming at everyone "it's wrong" - is just pathetic. If you'd actually analyse popular trailers you'll see this technique used constantly.

@@FilmmakerIQ I wasn't suggesting either the theory or practice of cutting one frame ahead is wrong - it was the original video that I felt was wrong (because it was, due to the KZfaq delay).

Nunya Bidness pretty ironic these days.

@@MaltevdH where's the lie...

Nunya Bidness depends on your definition of "freedom". The US today don't match MY personal definition of freedom a lot.

Internet download manager - and the original FPS is intact. This seems to be a common occurrence with h.264 encoding

I uploaded a ProRes file which had no audio delay. I don't think it matters what format you upload from, the KZfaq encoder adds it's own delay regardless

Actually in the case of interlace, there never were visible lines in that sense. The scanning gun moves so fast and dies out so quickly that you only think you see a full image. The interlaced lines only became visible when we started looking at that signal on a progressive screen.

Well about 5% of my comments on the previous video all echoed that sentiment - it was more that our minds got used to hearing sound later than seeing light, so the edit should reflect that. Problem is it assumes the human brain is some sort of insane chronometer capable of microsecond precision and then automatically adjusts for varying distance differentials. When really the truth is, our brains just sorta guess.

The speed of sound matters, though (yes, in the opposite way here), it's a real problem for large concerts.

You're misreading the situation. Nobody is talking about the delay between your speakers and your ears. We're talking about the delay between sound sources in real life and our ears. Most stuff we hear is between 0 and 200ms away from us (or between 0 and 68 metres). That means that our brain has been trained since early childhood (even ignoring evolution) that sound can never be early but can easily be late. It even expects that. However, we also have been trained to ignore the distance "behind" a TV or movie screen. Watching an explosion that 70 metres away in real life, we expect a 200ms audio delay. But watching an explosion that's 70 metres away from the camera on a TV screen that is 0.7 metres away, we expect...2ms delay.

@@HenryLoenwind but now you're talking about explosions which the human animal did not evolve to see... The only example of a large crash from those distances I can think of is lightning and thunder but if you're far enough where you can hear the differential then you're really not in any danger of it... The bigger point is that the human mind does not know what the differential between the sight and sound speed... There is no way of knowing. The only way we know today is because we have a separate millisecond chronometer that we can compare both phenomenons to... such a thing in the brain does not exist. Instead the mind just combines the concepts together if they fall into a relatively close period of time.

In premiere you right-click on the timeline bar at the top and select show audio units. Then you look for a wrench and switch it from audio units to milliseconds. I'm probably not describing it right but that's about how you do it..

@@FilmmakerIQ Ah, I see! It's listed as Audio Time Units." Thank you so much!

To me that's a recipe for a flam. The ITU chart demonstrates it... But some times that's desirable.

It's literally explained right there.... Did you watch the original video (the one before this one)?

Not that individual considering KZfaq does this to every single video on the site and really no one complains....

@@FilmmakerIQ That's because the majority never pays attention to it unless there's a huge deviation. I work within the dubbing industry and we do lipsync all day long. It becomes a part of your daily life to spot flaws in any media unfortunately.

Well even the ITU has a range of *undetectable* sync which is around 120ms... So you might be good but you're still human and you aren't _that good_ But if you like the audio 1 frame after the picture, that's what I've been advising so I think I misread you.

J and L cuts are more than a frame or two

Oh and if it is really important to get the exact sync - the odds are pretty good that the client will see it under the worst conditions ( as you mentioned ) I have seen horrible latency on the lowest quality settings.

Yes, in the original setting your brain compensates for the large distance, but when watching the recording, it only compensates for the distance to the monitor. I wonder if our brains would compensate for the distance "behind" the monitor if movies and TV didn't sync the audio to the picture but kept the correct audio delay for the distance between camera and action.

@@HenryLoenwind If for example the brain obtains the distance from the focal distance of the eye that would mean the perceived effect would differ between a monitor and a very large, distant screen such as in a cinema. I think the microphone characteristcs also made a big difference. The additional microphones supressed the building's massive echo much better than human ears or the in-camera microphone both due to characeristics and the location. The building in question btw. is the Dome of St. Blasien, see de.wikipedia.org/wiki/Datei:Dom_von_St.Blasien.png for an image of the interior. Choirce and orchestra were located in front rsp to the right of the altar in the centre and I was located on the balcony like 10m above ground behind the camera of that picture.

I use a program called internet download manager.