Software with Accurate Video and Audio Waveform Display?

[SaurusX]

I really have two questions:

1) How accurate is VideoReDo's video and audio waveform display synced up? When I use mkvtoolnix to add or subtract audio delay to a video it takes a certain amount of displacement before any difference at all shows in the GUI of VideoReDo. Therefore either AC3 tracks can only be moved in incremental steps of low granularity (I mean we're still talking milliseconds here) or VideoReDo is not exactly accurate

2) If VideoReDo is not accurate does anyone know of a program that is?

I'm just trying to sync video to audio from a DVD rip. Maybe I'm going overboard, but I'd like to get it as accurate as possible and this type of display seems ideal for doing so.

[Bernix]

in MKVtoolnix they suggest to get delay or other way round from VLC. If audio is ahead of imagepress f if audio is behind image press g.
It moves audio +/- 50ms this is 1,25 frame in 25fps or 2,5 frame in 50 fps video. Then use this number in Mkvtoolnix.
Or you can use NLE video editor program to adjust audio precisely but 50ms is not noticable.
50ms is similar to when somebody shout at you from 50 foots distance and you notice some delay. Then you can go in higher resolution. No problem set delay to 25ms, but I don't believe it is really noticeable.

Bernix

[pandy]

IMHO people will not realize +-100ms lipsync error (at leasts accordingly to few scientific research i was able to find).

[manono]

Yeah, and he's talking about milliseconds and AC3 audio has a minimum between chunks (is that the term?) of something like 40ms. Someone correct me if I'm wrong. If working with WAV audio, you can move it as few ms as you want.

[Bernix]

If this is issue, so anybody can generate waveform in Subtitle edit. Simply shift audio (not in Subtitle edit, the waveform is here just for subtitles) and then check how it was moved according to start, or there is grid 10 square in one second. So it can help too. But as Pandy mentioned and I said before, 40 ms isn't problem. And I believe if there is 40ms restriction for ac3 then Mkvtoolnix counting with it

Bernix

[Cornucopia]

According to this article: http://www.atsc.org/wp-content/uploads/pdf/audio_seminar/12%20-%20JONES%20-%20Audio%20...ion-Status.pdf, the threshold of mis-sync detection is +22ms/-30ms for ITU/EBU and +15ms/-45ms for ATSC, with threshold of unacceptability +90ms/-185ms for ITU/EBU (they don't list ATSC for that 2nd part). Note: +=audio earlier, -=audio later, note also this is end-to-end cumulative sync. Basically, 1/2 to 1 1/2 video frames to detect, beyond 2 1/2 to 3 video frames becomes unacceptable. This is true for most discerning users, though if you are a professional in an industry that relies on timing (TV/Film, Musicians, Dancer/Choreographers), detectability and unacceptability will be tighter than that.

Scott

[Bernix]

O.k.
then how theaters works? We can say 33cm is plus minus 1feet that is equal 1ms. If you are watching screen movie from 20 meters so it became unacceptable than?
I know there is some sound system, but there are theaters even stadiums that are bit bigger. So there are for example big screen of stage (concert) etc.
If you are in car cinema, 30 meters from screen (suppose audio is on same place) 30 m far from it, so you have to have awful experience. Glad I am not so sensitive

Bernix

[pandy]

Seem Dolby formal requirement are now significantly relaxed than few years ago... in fact some consumer HW from so called reputable company is unable to provide constant frame rate at the HDMI output... to be honest as a test engineer i see that whole industry care less every year on such things like timing, lipsync etc... Public broadcasters even distribute video with incorrect/wrong field dominance... (this is common for US game clips/ads broadcast in Europe, saw this in ZDF, ARD etc).
If delay is adjusted by modifying PTS then i believe it should be not a problem to set it with accuracy 1ms (usually clock is 90kHz).

Oh, forget to mention that ffmpeg can be easily used to observe audio/video time relation.

[manono]

I just checked and it's 32ms for AC3. You can adjust it in multiples of 32ms. For this kind of work I use AviSynth's Audiograph command. But I use WAV files with audio and video opened in Virtual Dub. You can delay the audio using the DelayAudio command. You can also do the same with AC3 audio, with the constraints mentioned earlier.

A=AviSource("MovieFinal.avi").ConvertToYUY2()
B=WAVSource("MovieFinal.wav")
#B=B.DelayAudio(0.0)
AudioDub(A,B)
AudioGraph(2)
Return Last

[Bernix]

But I think some delay of sound is desired.

OT - just checked muzzle velocity, before manono posted pictures, and now I know why is there idiom something like "Sometimes you hear the bullet" it should be "shot" instead of bullet. According table here
Bernix

[Cornucopia]

If you HAVE to err, it's better to err on audio being later, since that can naturally occur (as evidenced above) and so is kind of anticipated/expected.
But remember, a firing of a shot - in real life - has delay of audio vs. the video. A proper studio recording of a firing of a shot would/should take into account the expected delay in the audience venue and backtime the audio recording so that its arrival time is an equivalent facsimile to the real event in the same space. Studios do this, it's not just a theoretical/hypothetical exercise. (Similarly, 3D productions must properly gauge pixel shift distance to give an interoccular disparity that matches the venue/s - usually dci theatre or home viewing).
Use this knowledge to properly sync your A and V.

Scott

[Bernix]

Hi Scott,
to me seems best to shift audio at "acceptable" + milliseconds for first row, so it is "acceptable" - millisecond for farthest audience. If there is some outside cinema that has farthest 60 meters from screen so it is about 180 ms. According to your specs you provide, but there is only 52 or 60 ms threshold. So the threshold for 60 meters cinema should be 180ms respective 90ms.
Found this it is important from many points of view
But it is about actors and theatre plays.

Bernix