How Timecode Keeps Separate Files In Step

When picture and sound are recorded on different devices, something must state which moment of one corresponds to which moment of the other. Timecode is that statement, attached frame by frame.

Every frame gets an address

Timecode labels each frame with hours, minutes, seconds and a frame number. The label is a position on a timeline rather than a clock reading in the ordinary sense.

Because each frame carries a distinct address, alignment becomes a lookup rather than a judgment. Two files agree when their addresses agree.

This matters most where the picture gives no clue. A distant shot with no visible sync event is impossible to align by eye and trivial to align by address.

Devices drift apart without a shared reference

Each recorder counts time using its own oscillator, and no two oscillators run at identical rates. Over a long take the difference accumulates into a visible offset.

Drift is small per second and substantial per hour. A recording that starts perfectly aligned can end noticeably out of step, with lips leading or trailing the words.

The remedy is periodically resetting devices to a common reference, or feeding one continuously. Either way the shared reference is what holds the addresses honest.

Frame rate determines what an address means

An address is only meaningful alongside the rate it counts at. The same numbers describe different moments if two devices disagree about frames per second.

Broadcast frame rates that run slightly below whole numbers complicate this further. Counting whole frames against real seconds accumulates an error that must be corrected by skipping label numbers periodically.

That correction skips labels, not frames. No picture is lost, which is a persistent source of confusion when durations are computed by subtracting addresses.

Audio provides a fallback alignment

Where timecode is unavailable, alignment can be recovered from the sound itself. A camera microphone records the same acoustic event as the separate recorder, and matching those waveforms locates the offset.

This works well when the reference audio is usable and fails when it is not. Distant subjects, wind and heavy background noise all reduce the match's reliability.

A sharp transient serves the same purpose manually. A hand clap gives a single unmistakable spike in both recordings.

Alignment must survive the whole workflow

Timecode is metadata, and metadata is lost easily. Converting or trimming files with tools that ignore it produces media that will no longer align automatically.

Trimming is especially hazardous, since removing the beginning of a file changes where its addresses start unless the tool preserves the original values.

Checking that addresses survive an intermediate step is quicker than rebuilding alignment later. Once the labels are gone, the sound must be matched by hand.