AI music for product videos: the real sound often wins

A click, a pour, a zip, a lid coming off, a machine starting. These carry more conviction than a track does.

A viewer hearing the actual sound of a product gets evidence. A viewer hearing music gets production values.

On a demo, evidence wins. It also costs nothing, since you already recorded it.

Play the cut three ways: muted, real audio only, and with a bed. Most people never try the middle one.

If real audio only holds attention, stop there and keep the 20 credits.

Music earns its place where the product is silent, where the footage is a montage, or where there are gaps that feel dead.

The description box takes the sound in your own words, comma separated. Clean electronic production, synths, steady pulse is a complete brief.

Leave the lyrics box empty for an instrumental, which is what a product video almost always wants.

Pick the length your ad already is. The price is 20 credits at every one of them, so there is no reason to generate long and trim.

If there is a voice-over, name the absences: sparse, nothing in the midrange, low end and high percussion only.

If there is no voice-over, the track can carry more, and a sung line becomes an option rather than a conflict.

The 300 free credits with no card cover fifteen attempts, which is enough to find a sound your whole catalogue can share.

A product line that sounds like one line is doing brand work no individual ad does.

Generating a different track per product gives you forty unrelated ads that happen to share a logo.

So find a description that works, save it, and reuse it. Regenerating from the same words gives you a sibling track rather than a stranger.

That gives each SKU its own file while keeping the family resemblance, for 20 credits each.

It also removes a decision from every ad, which matters more than it sounds when you are shipping weekly.

Three saved descriptions, one calm, one energetic and one sparse, will cover almost everything a product catalogue needs.

Write them down somewhere the next person can find them, because the value is in the reuse rather than in any single track.

A phone speaker reproduces almost nothing below roughly 500 Hz, and most product videos are watched on one.

So the weight in the track is inaudible and the midrange is doing all the work, which is where any narration also lives.

That is why a bed that sounded rich on monitors turns muddy over a voice-over on a phone.

Check the finished ad on a phone at half volume held at arm's length before it ships.

If the words go unclear, generate a sparser track for another 20 credits rather than lowering the level until the music does nothing.

In the editor the controls are turn music on, pick the track and set how loud it sits. No ducking and no stems, so the description is the mix.

The order that works for a product video is footage, then words, then music. Nobody does it in that order the first time.

Music generated before the cut exists is a guess at a length and a pace that have not been decided yet.

Cut first, listen to what the ad already sounds like, then spend the 20 credits with something specific to ask for.

That sequence is why one attempt is usually enough here, and why a wrong answer costs a fifth of a finished ad.

One trap to avoid. A montage with a track over it is what people make when the product footage is not convincing on its own.

Music makes that ad feel more finished. It does nothing for the conversion, because the missing element was proof rather than polish.

The cheaper fix is better footage. Twenty seconds of hands and product on a phone, reused under every ad you make afterwards.

Four plain limits sit alongside. No editing after generation, no stems, no tempo slider, and 120 seconds is the ceiling.

A specific recognisable song for a launch is a rights negotiation, and no generator substitutes for it.

The track is generated for you rather than licensed from a catalogue. Confirm the usage terms that apply to your own account before running it behind paid spend.