RadioAIVT

How to voice track a radio show

Voice tracking is how one presenter can sound live on three stations at once, and how a small station can sound hosted at two in the morning. Done badly it is obvious within one link. Done well nobody ever knows. This is how it is actually done — the craft, not the software menus.

Written by a radio presenter with three decades on national air, who has voice tracked shows and built the tools that generate them.

What a voice track is

A voice track is one recorded presenter link, scheduled between two songs so the hour sounds live. You hear the end of one record and the intro of the next, you talk over the instrumental ramp, and the automation fires it at the right moment on air.

That is the whole idea. Everything else is craft: what you say, where it sits, and whether it sounds like you are in the room right now.

What you need

How many links per hour

Three to five is the normal range for a music hour. Below three, the hour drifts and stops feeling hosted. Above five, you are talking more than the format wants and the music suffers.

Place them where they earn their keep: over a strong intro, coming out of a set, into or out of a stopset. Not every song needs an introduction, and a link that exists only because there was a gap always sounds like one.

Riding the ramp

This is the skill that separates voice tracking that works from voice tracking that does not.

Every song has an intro — the instrumental bars before the vocal starts. That window is your runway. You start talking over the music and you finish exactly as the vocal comes in. Not two seconds early, leaving a hole. Not half a second late, stepping on the singer.

How to actually do it

  1. Learn the intro length first. Play the front of the song once. Nine seconds is nine seconds of writing, not ten.
  2. Write to that length, not to your idea. Say the thing that fits. If it does not fit, cut a word, not the pace.
  3. Land on the last beat before the vocal. The best links finish with a fraction of music still breathing underneath.
  4. Record it in one take, standing up. Voice tracking recorded sitting down sounds recorded sitting down.
  5. Listen back in context — outgoing song, your link, incoming song. Never in isolation. A link that sounds fine alone can be completely wrong between those two records.

What to say

The fastest way to sound automated is to read the log out loud. "That was X, and here comes Y" is not a link, it is a caption.

A real link does one of a few things:

One idea per link. Two ideas in nine seconds is two ideas nobody heard.

The mistakes that give it away

Talking about time you cannot be sure of

"This morning" recorded on Tuesday and aired on Thursday is the classic. So is a precise clock time in a log that can shift. Keep time references either exact and verified, or vague on purpose.

The same three openings

Every presenter has favourite ways into a link. In a live show they are spread across hours. Voice tracked in one sitting, they end up three songs apart and the pattern is audible.

Flat energy

You are in a quiet room with no audience and no adrenaline. The energy that felt fine in the chair sounds half a level down on air. Push slightly more than feels natural — and change it by daypart, because a midnight link recorded with morning energy is wrong in a way listeners feel without naming.

Weather and news that expire

Anything perishable has to be either genuinely live or left out.

Recording the whole show in one energy

Four hours tracked back to back drift into the same rhythm. Break it up, or deliberately reset between hours.

The test: listen to four hours in a row, not one link. Everything sounds convincing for twenty seconds. What exposes voice tracking is repetition and placement across a whole daypart.

How it works in the main systems

The principle is identical everywhere: the tool plays you the end of one song and the start of the next, records your link, and drops the file into the log at that position with the right cross-fade points.

Read your own software's documentation for the buttons. No guide replaces the tool you actually run.

When voice tracking by hand stops working

It is not the technique that fails. It is the arithmetic.

One four-hour show, five days a week, at four links an hour, is eighty voice tracks a week. Add the weekend and you pass a hundred. Cover the overnights too and you are in the thousands per month — and every one of them still needs a person in a chair.

Most stations get to the same place: voice tracking stays for the shows that carry the station, and the hours nobody was ever going to record stay silent. That is the gap RadioAIVT was built for — it generates the links for those hours from your own log and delivers them into your automation, while your presenters keep the shows that matter.

If you want the numbers side by side, we wrote it out: hiring a voice tracker vs AI voice tracks. For how the generated version works, see AI voice tracking for radio.

Hear a generated hour Listen to the voices →

Common questions

How long should a voice track be?

As long as the intro allows. Most links sit between five and fifteen seconds. The constraint is the ramp, not your script.

Can listeners tell a show is voice tracked?

Not from one link, if it is placed well. Over hours, what exposes it is repetition, flat energy and time references that do not hold up.

How far ahead can you voice track?

The further ahead, the more careful you have to be with anything perishable. A day ahead is comfortable. A week ahead means cutting almost every time reference.

Do I need expensive equipment?

No. A decent microphone in a quiet, treated space beats an expensive microphone in a live room every time.