6 min read
Automatic audio sync: how the software works and why it sometimes refuses
Auto sync is not magic. It is correlation, and correlation needs something to correlate with.
Press sync, wait, and either the whole day lines up or a list of failures appears with no explanation. Understanding what the algorithm looks for turns that list into something you can fix in minutes.
What waveform sync actually compares
The software takes the camera's scratch audio and the recorder's audio, reduces both to an energy envelope, and slides one against the other looking for the offset with the highest correlation. It does not understand speech. It matches shapes.
Why it fails
- The camera track is too quiet, too distant or drowned in wind, so there is no shared shape.
- The scene is almost silent, giving the algorithm nothing distinctive to lock onto.
- The recorder track is a poly file and the tool correlates against the wrong channel.
- Sample rates differ, so the two envelopes stretch relative to each other.
Timecode sync is a different mechanism
Timecode matching reads the bext time reference from the sound file and the camera metadata and aligns numbers, not waveforms. It is instant and scales to thousands of clips, but it inherits any frame rate mismatch, which is why 23.976 material read as 24 drifts about 3.6 seconds per hour.
Prepare files and most failures disappear
- Split poly WAVs into named mono stems so the tool syncs against the boom, not a mix channel.
- Confirm every sound file is the same sample rate as the timeline before importing.
- Check one file per card for timecode and frame rate flags.
- Keep the camera scratch track enabled even when it sounds bad, it is the correlation reference.
The inspector here reads sample rate, bit depth, timecode and channel names locally, and the splitter produces the named stems, so the sync pass starts from clean material.