Voice tracking: how pre-recorded radio sounds live

Voice tracking is the practice of recording a presenter's links separately from the music, and having automation play them in the right places later. Done properly, a listener cannot tell. Done badly, it is the most obvious thing on the radio.

It is how a station with three presenters covers twenty-four hours, how a breakfast show exists on a Saturday, and how a volunteer with a day job hosts a weekday afternoon programme.

What a link is

The unit of voice tracking is a link: the bit between two songs. Typically ten to thirty seconds — a back-announce of what just played, something in the middle, a forward-reference to what is next.

A presenter recording a three-hour show does not record three hours. They record twelve to eighteen links, each under half a minute, and automation places them. The whole show takes fifteen or twenty minutes to produce.

Why it sounds live when it works

Three things do the work, and only one of them is technical.

The segue. In a live show, the presenter talks over the intro of the next song and the outro of the last one. Voice tracking that plays a link in silence between two songs sounds like an announcement, not a show — the dead air on either side is the tell. Good voice tracking overlaps: the link starts under the end of the outgoing track and ends as the vocal of the next one begins.

The references. "Coming up after this" works. "It's twenty past three" works only if the show airs at twenty past three, and stations that re-air voice-tracked shows learn this the hard way. Time references, weather and news references, and anything about what happened this morning are all commitments to a broadcast slot.

The energy. Recording alone into a microphone with no music playing produces a flat read. Presenters who monitor the outgoing and incoming tracks while they record sound like they are on the radio, because they are reacting to something. Presenters who read a script in silence sound like they are reading a script in silence.

What it is not

Voice tracking is not a robot, an AI voice, or a station with no people in it. It is a human presenter making the same editorial choices they would make live, just at a different time of day.

It is also not the same as a pre-recorded show, which is a single continuous file with music baked in. The difference matters:

For anything music-led, voice tracking is the better shape. Pre-recorded files are right for a documentary, a mix show where the mix is the point, or a syndicated programme you did not make.

Where it fits in automation

A link can be attached to two different things, and choosing correctly is most of what there is to learn.

Attached to a date and time — this link belongs in Tuesday's 3pm show, once. That is normal show production.

Attached to a playlist — play this introduction before that song, every time it comes round. This is how a station builds character over years: a presenter records forty song-specific intros, and they surface whenever the songs do, in whatever show pulls them.

The second kind raises a question the first does not: should it also air in automated clockwheel hours? Usually yes — a song pulled into a clock slot introduced exactly as it would be in a playlist show is the whole point. Occasionally no, when the link references a specific programme. In RadioHost Studio this is a per-link setting, and a missing link is nearly always that setting rather than a fault.

Clocks also carry an explicit talk break slot, which pulls the next link queued for that show. If nothing is queued, the clock steps over it and moves on. That is what makes voice tracking optional per hour without redesigning the hour.

Recording well

Monitor the music. Record with the outgoing and incoming tracks audible, so the read is timed against something real.

Do not over-produce. Voice tracking tempts people towards perfect takes. Radio is not perfect takes; it is a person talking. Second take, keep it, move on.

Keep links short in daytime. Fifteen seconds of something is better than forty seconds of anything.

Record in one sitting. A show recorded across three days has three different microphone positions and three different moods in it, and that is audible even when nothing is technically wrong.

Batch by show, not by day. Presenters who record Monday-to-Friday's links in one Sunday session sound consistent all week and spend less time doing it.

Where it stops working

Voice tracking cannot do live. It cannot take a phone call, react to a goal, cover a weather emergency or acknowledge that something has happened. A station that voice-tracks everything is a station with no way to be present on the day being present matters.

The usual shape for a small station is voice tracking as the floor, not the ceiling: it makes every hour sound hosted, and it frees the human time for the hours where being live is the point.

Frequently asked questions

Is voice tracking legal? Do I have to tell listeners?

There is no general obligation to announce that a show is voice-tracked, and stations have done it for decades without disclosure. What is not acceptable is claiming to be live when you are not — inviting calls to a show nobody is monitoring, or referencing a live event you have not seen.

How long does it take to voice-track a three-hour show?

Fifteen to twenty-five minutes once a presenter is used to it. The music is not recorded, only the links, so the show length has little bearing on the production time.

Can voice tracking be done from a phone or a laptop away from the studio?

Yes, and it is one of the main advantages — a presenter with a decent microphone and somewhere quiet can produce a week of shows from home. Consistency of microphone and room matters more than the equipment being expensive.

What is the difference between voice tracking and an AI radio host?

Voice tracking is a human presenter recording at a different time. An AI host is synthesised speech. They solve different problems: one is scheduling, the other is whether there is a presenter at all.

Does voice tracking affect royalty reporting?

Not adversely — it improves it. Because the music is pulled from the library at play time, every track is logged as an individual spin, which is what royalty reports need. A pre-recorded file with music baked in has to be reported from a manual cue sheet.