1

If I have an interview or lecture style video and the audio is out of sync, by how much can it be out of sync before the average person would easily notice?

I'm currently trying to create an HTML5 player that sometimes has the video and audio in separate elements, meaning the video and the audio buffer and play at separate times. I've developed a decent code that keeps them synced (it's a harder problem then it should be). The current code pauses the video and the audio elements if they are off by 0.1 seconds, adjusts the time of the video element to the audio element, waits for oncanplay event, then plays both. My question is whether 0.1 seconds off in audio/video sync is too much. Would a common user notice a 0.1 seconds sync disparity? What's the highest I can go before the common user might notice the lips don't quite match the voice? Does any large organization have an official standard on this?

Improve this question

1 Answer 1

Reset to default
5

There are, in fact, several standards on audio/video or lip sync. The problem is, they can't all agree on what the acceptable error is. What is well understood is that if the audio lags the video it is less noticeable, because that is natural. Something that is further away will be seen before it is heard. In the list below, a negative number is the audio lagging the video and positive is the audio leading the video.

ITU BT.1359 was based off studies done with average viewers, but relied on significantly older technology. The ATSC and EBU recommendations rely on expert opinions. The methodologies don't give you a hint at which one is right. I've worked at facilities that used -2 frames to 1 frame (roughly -33ms to 16ms at 60fps) as an easy to remember guideline with very few viewer complaints about lip sync.

Even with the disagreement, 0.1s (100ms) of error will almost always be noticeable, unfortunately.

Improve this answer
5
  • 1
    Great, this helps a lot. I think I'll update my margin to -100 ms to 45 ms, per the ITU BT.1359. This conveniently also answers if positive is more noticeable than negative. I'm tentatively giving you the selection, but might change it if a better comes along. Not that your answer isn't good. You've earned the selection. It's just that I don't usually like to give it so early.
    – user3643
    Commented Oct 23, 2018 at 3:42
  • 1
    given that at 25fps each picture is displayed for 40ms there's a +/- 20ms margin for sync even for sound recorded with the picture. Say the shutter opens at the start of the frame, while audio of course records continuously. At the beginning of the frame the picture will be in sync with audio, and at the end the picture is 40ms behind. So one wonders how you can possibly be 15ms out of sync.
    – stib
    Commented Oct 24, 2018 at 0:11
  • Keep in mind, the ATSC is US-based organization and the recommended practice was developed for a system that is predominantly 30/60 fps. At those frame rates, one field/frame is roughly 16ms. It basically says that if the audio leads the video by even one frame, it is noticeable and that is consistent with what I've run into. The European (EBU) standard matches what you said about a 25 fps world.
    – Michael Liebman
    Commented Oct 24, 2018 at 3:00
  • 1
    The SMPTE handout appears to have been removed from their site - archive.org has a copy at web.archive.org/web/20150328054944/https://www.smpte.org/sites/…
    – Peter Barton
    Commented Oct 22, 2020 at 22:54
  • @PeterBarton I edited the link, thanks for finding the archive. SMPTE recently went through a branding refresh, so I guess that swept away a bunch of older stuff.
    – Michael Liebman
    Commented Oct 23, 2020 at 2:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.