Several factors affect audio lip sync with video under various circumstances.
Why do I hear audio a little early (or a little late) when I'm stepping through frame by frame?
Audio scrubbing is not a native browser function, so we've had to implement it ourselves. In order to make the sound of audio scrubbing/stepping more pleasant (and to match the behavior in Digital Content Creation tools), we introduce tiny crossfades at each end of the little bursts of audio associated with a video frame. This "pulls in" audio from surrounding frames, which can make it appear that the audio is starting early or ending late. However, the audio is actually in sync, as shown in the waveform seen in the background of the timeline when clips are under three minutes long.
When I download the transcoded version of my file and import it into a DAW, I can see silence inserted at the beginning of my clip. Doesn't this slip the audio out of sync?
Due to a quirk of the way browsers play MP4 files with AAC audio, starting the audio exactly at the beginning can cause "ghost" frames in the playback. To eliminate this, we intentionally introduce about 46 milliseconds of silence at the beginning when processing the files you upload. These files will be in sync during normal playback.
For technical details about why we insert the silence, please visit:
Why does audio start at the beginning when I added it in a later frame in Maya?
Currently we start all audio at the beginning of the video. We will support playblasts with audio that is meant to start elsewhere in the timeline in an upcoming release.