Audio Sync Comparison for Various Render Methods

john_dennis wrote on 4/21/2018, 8:39 PM

In this post there was a reference to troublesome audio sync with the lip movement. I decided to create a video from my own camera to use to test my various delivery methods. I recorded the butt of a drumstick hitting a piece of metal with a clock running in the background for general reference. The recording was 1280x720-29.97p AVC + PCM in a MOV wrapper. I found the Vegas2Handbrake encodes produced audio that leads the video by 48ms for a single midstream sample. Other popular encoding methods that I tested were surprisingly accurate with the exception of Mainconcept AVC Internet which led by a small amount. Though unlikely to cause a complaint, I would prefer it to lag if there was a variation. Since this is forum about making video...

Change Log

2018-04-25 Replaced video with improved titling.

Back to post

Comments

ottor wrote on 6/24/2018, 10:02 PM

It looks to me as if there's a problem decoding aac audio in Vegas, at least in versions 12, 13, 14 (I don't have 15). The results vary depending on the source encoder.

Method:

I rendered a 1kHz tone to an MXF file, 24bit linear audio using Vegas 14.

I then rendered the MXF to an MP4 file with 512kb/s AAC audio in Vegas 14 using Mainconcept AVC codec.

I also rendered the MXF to an MP4 file with 512kb/s AAC audio using an Elemental Server encoder.

I loaded all 3 files into Vegas, and the Elemental Server file had it's audio offset by about 25 milliseconds! The Mainconcept MP4 also has a slight offset.

(As a check I then rendered the Elemental MP4 back to wav using the Elemental and the output file is back in sync with the original MXF. So it looks like Vegas' decoding is the culprit here.

Does anyone have any other AAC encoders they can check against?

Otto

Musicvid wrote on 6/25/2018, 9:55 AM

I have access to FDK-AAC, but not with Sony mxf as the source. If you can provide XAVC-I, I will be able to test.

john_dennis wrote on 6/25/2018, 9:57 AM

Do you use Elemental Server in your day-to-day workflow? You could try Sony AVC/MVC AAC Audio Only up to 320Kbps and multiplex it. It's right on when put back on the timeline of the original project after going through MXF.

Musicvid wrote on 6/25/2018, 10:00 AM

Also, it's important to conduct listening tests on many players. VLC, WMP, QT all use their own offsets, and audio sync is the first thing to go under load from a challenging video bitrate, x2 for Quicktime.

ottor wrote on 6/25/2018, 5:18 PM

Hi John,

Yes, Elemental Server (I use 3 of them) is the workflow. I am using Elemental to create 20mb/s mp4 files with embedded aac audio from many different source formats (ProRes, MXF, ts, mp4).

Occasionally I receive an additional language file that I need to sync to the aac ausio of a previously encoded mp4 and output it as a wav file. I was doing this in Vegas, but as I have now discovered, Vegas incorrectly decodes the aac audio in the mp4, clipping about 25ms from the start, so I can't use Vegas for this anymore unless I first transcode the mp4's aac audio to a wav using Elemental.

BTW, Resolve & Premiere correctly decode the aac from Elemental's mp4, so it looks like Vegas is the one with a problem.

ottor wrote on 6/25/2018, 7:00 PM

Here's a short file encoded in handbrake. It starts with 1 frame of 1kHz tone and Colour bars, then silence/black until 1 second, tone/bars again until 2 seconds, silence/black until 3 seconds, tone and bars until 5 seconds. If I play the file in MPC-HC, VLC or Windows Media Player I can see/hear the bars/tone at the 1st frame.

If I load it into Vegas the colour bars for frame 1 are there, but the tone is not and the whole audio is shifted to the left... and Vegas is not alone. Switch 4.1 and Quicktime 7.7.9 also fail to play the audio of the 1st frame.

vbencev wrote on 8/26/2018, 4:47 PM

Yes wwag, that is an internal delay that vbr (non-interleaved) material always needs, and it is invoked in the file headers, even old-school mpeg-2,

It is rarely exposed, but drifts a little, and can be tweaked in yamb/mp4box if you are a perfectionist who also is compelled to drive himself batty.

How much is the usual value of this delay in case of an mpeg-2 with ac3? Inoticed when I reloaded exported mpeg-2 to Vegas, with ac3 audio there is +5ms delay relative to the project from which it was rendered.