Consistent audio/video desync (150-200ms); capture or post?
 

Hi all,
I have about a 150-200 ms desync where my audio plays ahead of my video.

• consistent desync from beginning to end
• default timing settings in VDub (for example, Audio Latency Determination: Automatic; Number of audio blocks to use at start: 30)

1. Is it better to fix this in capture or in post?
2. How can I get a precise value? Do I use something like premiere to figure out the offset? I don't want to do the process of capturing different videos of different values to try to figure out what the perfect number is. (Or is that the best way?)


Capture Workflow
- JVC VCR (SR-V101US)
- Cypress AVT-8129 Composite TBC
- ATI TV Wonder HD 600

PC SETUP
- AMD 8Core 3.1Ghz, 8 Gb RAM, NVIDIA GeForce GTS 250)
- Windows7 Professional, Service Pack 1 (32-bit OS)
- NOT connected to the internet
- NVIDIA 342.01 Win7 for Video Card
- ATI Catalyst Media Center Install Manager through ISO + Gizmo Central (this installed the C++ redistributable, USB Video Driver)
- VirtualDub1.9.3 + Filters

Do i understand correctly that this is a steady 5 to 7 frame latency where in the video lags the audio?

How are you evaluating/measuring the latency? If by ear (listening to the audio while watching the video) please describe the audio and video (if separate files) and the detailed signal paths.

I'm just watching the video in virtual dub itself and doing brute force trying different timing options in the vdub interleaving option menu and trying audio skew correction. So far a 17-18ms skew looks correct for me.

I think there might actually be a 2-3 ms drift between the beginning of my video vs. one hour into the video, it's hard to tell exactly when I stare at the human lips and audio long enough.|

Is there a better process for this?

NTSC video is nominally 30 FPS (60 fields) - well 29.97 bit 30 is close enough.

Thus one NTSC video field is ~16.7 ms. FWIW, sound moves at about 1125 feet per second, so the 16.7 ms is equivalent to moving the audio recorder about 19 feet closer/further from the source.

In nature sound always lags visual due to propagation speed differences between sound and light waves so audio leading videos is unnatural and noticeable. Sound lagging a similar amount is not as readily noticed.

However, electronically speaking video frame processing and display takes longer that audio processing. (A full frame TBC has to receive a complete frame before it can spit it out. Sound usually bypasses the TBC.) Thus the audio output of a PC sound card may lead the audio (and video) in a HDMI stream sent from a the same PC to a TV set.

A secondary issue is that the clocks in separate video and audio A/D devices might not be precisely accurate although this was more of an issue in the past. This can cause drift.

The easiest way to ensure audio and video are and remain in precise sync is to use a quality capture device and software that puts both the audio and video into a single file. Alternatively learn to use the separate tools to compensate for differences.

And keep in mind what counts is the audio sync in the final product/deliverable be it DVD, Blu-Ray, or MP4 when played on the client/customers device.

I understand the notes in your reply. Thanks for the summary -- so far, I think I have fairly good quality capture set since it was a full set that I purchased from a person on this forum.

I still haven't found the best settings for virtualdub though, I've done several test captures, but nothing perfect so far.

Playing with anything away from default in Capture Timing Options seems to make things worse though. :laugh: