Prove the failure in the finished output first.
Make a short clap or visible keypress recording at the start and end of a representative session. Measure whether the sound-picture offset is constant, changes over time, or affects only one source.
What it may mean
A stable source delay, device timestamp behavior, duplicate capture path, sample-rate mismatch or accumulating device drift may be involved.
What it does not prove
One random playback check does not prove the offset is stable. A fixed sync offset cannot repair drift that keeps changing during the session.
Match the evidence before changing settings
A stable source delay may be measurable and correctable with one documented offset.
NEXT TESTAdjust only the affected source, then repeat the same visible cue.
Clock or timestamp drift is more likely than one fixed delay.
NEXT TESTCompare device timestamp behavior and sample rates before adding a larger fixed offset.
The capture path may add a source-specific delay that the microphone does not share.
NEXT TESTMeasure the card's picture-to-audio offset separately from microphone-to-camera sync.
One conservative free test
If one source has a stable measured offset, document its current Sync Offset and adjust only that source by the measured direction and amount. Repeat the same start-and-end cue before touching device timestamps or sample rates.
Run a repeatable audio preflight →Preserve the rollback
Write down the original Sync Offset, timestamp option and sample rate. Restore each value when the matching source does not improve or when drift replaces a fixed offset.
Technical reference: OBS Project: Audio Sources ↗
Common questions
01How do I know what OBS sync offset to use?+
Measure a repeatable visible cue in the finished file. Do not guess from the live preview; write down the original value and change only the affected source.
02Why does audio drift farther out of sync over time?+
Growing offset suggests clocks, timestamps, buffering or sample-rate behavior rather than one constant delay. A fixed offset may align one moment while leaving the end wrong.
Did the relevant counters improve?
For dropped frames, encoding lag or rendering lag, compare a completed log from before and after your change. For audio and black-screen problems, inspect the recording itself; counters cannot establish picture or sound quality.
Compare logs free →