Audio/video sync, timecode, deliverables, stems and broadcast specs.

A mix that sounds good but arrives out of sync, without stems or outside the channel’s range is not a delivery: it is a problem that gets sent back. The final stage of audio for video depends on verifiable details—a shared clock, a properly named package and numbers within specification—and none requires more talent, just a method.

This chapter covers Audio/video sync, Timecode, Deliverables, Stems and Broadcast specs: how synchronisation is maintained, how time is numbered, which files are sent and what broadcast destinations require.

Audio/video sync: keeping picture and sound together

Audio/video sync fails for ordinary reasons. A phone recording with a variable frame rate, a 44.1 kHz session against a 25 fps video, a converter with latency or a clip retiming that nobody checked can be enough for the audio to end up half a second ahead of the picture by the end of the piece.

There are three rules of defence. First, a single reference: either timecode on every track or sync points marked in the first and last relevant shots. Second, one sample rate, 48 kHz, set when the project is created. Third, check alignment at the beginning, middle and end, because drift accumulates without warning.

When an offset appears, correct it once at the source: move the complete track to the correct point and check the rest of the session against that correction. Moving clips one by one by hand produces the familiar effect of audio moving in and out of sync during the shot.

It is also useful to distinguish source synchronisation from playback synchronisation. Material can be perfectly aligned in the session yet sound late on a television with picture processing: that is no longer an editing problem but a playback-chain problem, solved in the player or receiver settings. Checking on two different devices—one with speakers, another with headphones—separates a working problem from a viewing problem.

Timecode: the project’s shared clock

Timecode is a numbering system of hours, minutes, seconds and frames that lets audio and video refer to the same instant. With it, a point in the mix and a point in the edit are the same place, which lets you locate review notes, place effects and reopen the session months later without having to reinterpret anything.

Working with timecode means making three decisions at the start: the project’s frame rate, which should match the source material; the numbering format, with or without skipped frame numbers in the 29.97 and 59.94 fps standards; and the start time, usually a whole hour so there are no negative hours. These concepts are explained in the guide to synchronisation with genlock and timecode.

In the session, timecode appears in the transport display and the editing ruler, and is used for spotting: placing a clip exactly at its original position rather than dragging it. It is also carried in Broadcast WAV, which stores timecode inside the file, so that a standalone file knows where it belongs.

Stems: delivering the mix in blocks

Stems are complete submixes that reconstruct the master when summed: dialogue, effects, foley, music and ambience are the usual groups. They are neither individual tracks nor versions with effects stripped away, but blocks already mixed within their own family.

They are delivered for a practical reason: any later correction—lowering the music in a scene, replacing a line or adapting the piece to another language—is made on the relevant block without redoing the whole mix. For this to work, all stems share the same start, length and bus processing, and none has its own normalisation: if each block arrives at whatever level its exporter chose, their sum does not reproduce the master.

If you work as a team, review how to exchange stems between DAWs before sending: the same sample rate, the same time references and an included check mix so that the sum can be verified.

A quick check catches most problems: sum all the stems and compare the result with the exported master. If the sum does not match—more bass, brighter or simply louder—one of the blocks has its own bus processing or different gain. Doing that comparison just once before sending saves the round of corrections in which the recipient tries to guess which stem has changed.

Deliverables: what to deliver and in which format

Deliverables are the complete package, not a single file. The usual set includes the stereo master, stems, an M&E mix without dialogue for dubbing, a reference video with visible timecode, the level track and a sheet with information about the piece: duration, format, channel configuration and the loudness at which it was delivered.

Keep the format simple: uncompressed PCM at 48 kHz and 24 bits, in Broadcast WAV when the destination accepts it, with embedded timecode. The filename follows a single convention—project, version, piece and channel—to avoid the familiar question of which of the four copies is the right one; the full organisational approach is in the guide to naming, folders and versions.

Document the decisions too: whether the delivery includes a clean version without music or with different levels, whether there are alternative tracks and what changes between them. A package that makes sense without calling its sender saves an entire round of correction.

Broadcast specs: requirements that are not negotiable

Broadcast specs are the technical conditions set by the destination that determine whether the piece passes or is returned. The most closely watched requirement is loudness, measured with a meter that meets the standard’s specification: -23 integrated LUFS under the European EBU R 128 framework and -24 LKFS under the ATSC A/85 recommended practice for US digital television.

Added to that are the true-peak ceiling, usually below the system’s maximum level to avoid peaks during encoding; the channel configuration—stereo, 5.1 or the map the channel requests—and the delivery container and encoding format. Many destinations also check for abnormal silences, a centred signal and the absence of distortion.

Other details can cause rejection too: headers with seconds of silence, tails longer than requested, a duplicated channel instead of genuine stereo, inverted phase on one channel and tracks misaligned with the reference picture. None appears on the loudness meter, so check them separately: listen in mono, inspect the ends of the file and review channel configuration before uploading anything.

Quality control uses the same meter the channel will use, along with a complete listen on headphones and speakers. When that check is skipped, the return comes with the review bill attached.

The complete delivery process

The order that reduces incidents is: confirm the destination’s specification before mixing; finish the mix with the meter visible; export the master and stems from the same session without changing intermediate gains; generate the reference video with timecode; check synchronisation and level on the exported material—not on the session material; name and package everything using the agreed convention; and archive the project, stems and master.

Always check the final file. It is the only moment when you validate what actually arrives, and the quickest of all: five minutes of listening avoids the round of corrections that costs a week.

Next step

Delivery completes the linear workflow. If the destination asks for more than two channels or an experience that surrounds the listener, the next step is surround, immersive audio, spatial audio and audio for installations.