Leave the lyrics empty and the track comes back wordless

Leave the lyric field empty and the track comes back with no vocal on it. Then describe the sound in ordinary language.

Warm acoustic guitar, slow, unhurried. Or dry electronic pulse, minimal, modern, a little cold. Or soft piano, sparse, nothing dramatic.

Reach for texture words rather than genre labels. Warm, dry, sparse, soft, steady, bright, muted. Those describe the space a track occupies, and the space is what you are buying.

Genre labels import an entire arrangement, complete with a build and a drop you cannot use under a sentence.

The safest brief for an ad bed is three words. Steady, low, uneventful. Each attempt is 20 credits, so a wrong brief costs less than the meeting about the brief.

The human ear treats voice as the most important object in any mix. That is wiring rather than preference.

So when a sung vocal and a spoken read arrive together, attention has to split. The viewer follows neither properly and concludes the ad is hard work.

It is physical as well as attentional. Sung vocals and speech occupy a similar frequency range, so they mask each other whatever you do with the level.

Lower the music and push it back, and the mud stays, because the mud is structural. Removing the vocal removes the problem rather than managing it.

That is why almost every voice-over ad you admire has an instrumental under it, and why leaving the lyric field empty is the default here.

In the editor, music is a switch, a track and a level. That is the whole surface on purpose. For a bed under speech there is one decision that changes the outcome.

Set it on a phone speaker, not on headphones. Headphones flatter everything and lead people to set music too high. That is the most common reason a well-written ad tests badly.

Play the ad on a phone at half volume. If you have to concentrate to follow the words, take the level down and play it again.

One more test costs nothing. Listen with the music off, then with it on. If the version with music is not obviously better, ship the silent one.

A rough starting point. Quiet enough that you would not notice the track if you were not listening for it. Then take it down one more notch before you export.

One number makes this concrete. Two exports of the same ad, one with the bed and one without, cost 20 credits between them. That settles the argument better than any opinion in the room.

It does not sync to your cuts. The track has no knowledge of your edit points, so a bed plays under a section rather than hitting a word.

It is not edited or shortened by hand. If the arrangement is wrong, describe it again and generate another for 20 credits. That takes less time than opening an arrangement view.

It does not duck itself. There is no automatic level following your speech. You set one level and it holds.

Which is exactly why setting it lower than feels right in a quiet room is the correct instinct.

And it will not make a flat ad interesting. If you push the music up to give the ad energy, the problem is upstream in the read or the script. Ship the ad silent and rewrite the first sentence.

20 credits per track, whatever its length. A $15 top-up of 1,000 credits is fifty beds.

Basic is $39.99 a month for 2,500 credits. Premium is $79.99 for 5,000. Both leave room to generate four candidates per ad and never think about it again.

Set that against the batch at 100 credits and the export at 20 credits per finished minute. Music is the cheapest decision in the product.

Four candidates is 80 credits, less than one export of a four minute cut. So the honest advice is to stop reading about music and generate the four.

The failure mode is not expensive. A bed you dislike costs 20 credits and four minutes, which is less than the time it takes to explain to somebody why you dislike it.

Then it becomes a switch, a track and a level under your read, in the same editor that cut the ad. No export, no import, and no second library to keep paying for.