Audio Post for Episodic & Vertical Series.

A season is a batch, not a project. Serialised vertical work typically ships sixty to a hundred episodes of one to two minutes, off a shoot that took seven to ten days, and post is counted in days rather than months. That shape is much closer to a campaign than to a film — a set of deliverables, each measured on its own, against a fixed date — which is the way this studio has worked since 2018.

Dialogue is the spine, and in a two-minute episode there is nowhere to hide it. Every line carries plot; there is no establishing sequence to bury a problem under. Fast shoots produce the same set of problems every time — room tone that changes between setups, air conditioning, a boom that drifted, one performer several dB under the other. The decision that actually matters is which lines can be saved and which need replacing, and that is a judgement, not a plugin. Restoration tools take out the noise and some of the performance with it; a person decides how much to take.

⚠️ The M&E is the file that decides whether the series can travel. An M&E is the finished mix with the dialogue removed — music and effects only — and it is what a dubbing house needs to put another language over the picture. Without one, the series cannot be sold into a territory that does not speak its language. It has to match the final mix exactly in timing, level balance and edits, or the dub drifts against the cut. It is routinely missing, because the first release never needed it — and it is far cheaper to build alongside the mix than to reconstruct afterwards out of a finished stereo file.

Every episode is measured on its own, because integrated loudness is an average across a whole duration. An episode with a quiet cold open and one without will read differently off the same session and the same settings — measurement follows ITU-R BS.1770, the standard underneath every loudness figure in broadcast and streaming. A batch is not one file repeated a hundred times; it is a hundred measurements against one target, which is why the level convention is agreed before the first mix rather than discovered at delivery. Delivery is 48 kHz, 24-bit, the film and broadcast standard, with stems sample-aligned to the master so a later change drops in frame-accurate.

The phone is the venue, and that changes the mix rather than the master. A phone loudspeaker puts out almost nothing below roughly 400 Hz, so weight that lives in the sub arrives thin on the device the audience is actually holding. The answer is not to brighten everything — it is to carry the low end’s intention in the mids, where a small driver can reproduce it. Dialogue also sits further above music and effects than a cinema mix would tolerate, and the band that decides intelligibility runs roughly 2 to 4 kHz, where the consonants live — so a music bed written through it will fight every line underneath, and riding a fader afterwards does not fully recover what was arranged wrong.

⚠️ On delivery specifications, one thing is worth saying plainly: the vertical platforms do not publish them. We went looking. The major apps operate submission portals rather than public technical documents, and the loudness figures circulating online for them trace back to vendor compilations rather than to anything a platform issued. The practical consequence is that the numbers binding you are the ones in your contract, not the ones on a blog — so ask for them in writing before the mix, not after a rejection. Where a platform states nothing, the published standards are still the only defensible anchors: EBU R 128 sets −23 LUFS with a −1 dBTP ceiling for European broadcast, ATSC A/85 sets −24 LKFS with −2 dBTP for the US, and streaming platforms normalise toward roughly −14 LUFS on playback rather than rejecting anything. Pick one, write it down, and hold the whole season to it.