Choosing a presenter and a voice for your ad

Of all the choices on an ad form, the presenter is the one people spend the longest on and have the least guidance for. Who should be on screen? Does the voice matter as much as the face? Is the more expensive quality level worth it? This article walks through those decisions in the order that makes each one easier, using the options that exist on Clipelle's ad form.

What an AI presenter is, in plain terms

In Clipelle a presenter is a portrait look from a library: a person, framed vertically, whose face and mouth are animated to speak the script that was written for your ad. You pick a look, a language, an accent, a voice and a speaking speed. You do not upload footage of yourself, and you do not train anything; the presenter is ready to speak the moment the script exists.

The presenter carries the whole ad. The product clip generated from your photo appears over them at fixed moments, and the captions follow the words, but the picture and the voice are the presenter's from the first second to the last. So the choice is not decoration. It is who your customer meets.

Start with the audience, not with yourself

The most common mistake is picking the presenter you like rather than the one your buyer would listen to. Ask three questions:

1. Who buys this product? Age, situation, what they are doing when they see the ad.
2. Who would they take a recommendation from? Often someone like them, sometimes someone slightly ahead of them.
3. What does the product promise? Calm, fun, precision, comfort. The presenter should look like they could plausibly say that sentence.

Then filter. Clipelle shows portrait looks only, because the ad is 9:16 and a landscape look would sit in a small strip in the middle of a tall frame. A gender filter narrows the library; the curated looks come first and a default is already selected, so doing nothing also produces a sensible result.

Standard or Premium

Clipelle offers two presenter quality levels, and the form shows the estimate for your chosen length at each.

Standard is recommended and preselected. It is the baseline for most ads: natural enough for a talking head in a feed, and it costs one credit per second of video, which is how the whole credit system is defined.

Premium adds more natural head and hand movement and the best lip-sync, and costs more per second. Choose it when the face is the whole picture: a tight talking-head ad for a 30 or 45-second slot where any lag between mouth and audio would be the thing people notice. For a 10 or 20-second test where the question is "does this hook work?", Standard answers the question at a lower cost. The Standard or Premium guide goes through the comparison with the costs side by side.

A good habit: test on Standard, then re-run the winning script on Premium if the ad is going to run for a while.

Language first, then accent

One language choice on the form drives three things: the script is written in it, the captions show it, and the voice speaks it. Pick the language of the market, not the language of your office.

Accents are offered where voices exist for them, and this is a choice worth making deliberately. A presenter speaking the viewer's own variety of a language sounds local; a "neutral" voice sounds like it is from nowhere, which in practice means it sounds like an ad. If you sell in one country, use that country's accent. If you sell in several, make a version per market, which Clipelle makes cheap enough to do.

The voice

Clipelle filters voices on the language, the accent and the gender of the presenter you picked, and preselects the voice that belongs to that look. You can preview voices one at a time.

The preselected voice usually fits best, because it is the one the look was recorded with. Change it when something specific is wrong, not for variety. What to listen for in a preview:

• Energy against the product. A sleep product and a party speaker should not share a voice.
• Pace. Some voices read fast at normal speed; with captions highlighting each word, a fast voice can feel rushed on a 10-second ad.
• Warmth. Does it sound like a recommendation or an announcement?

Before anything is paid for, Clipelle checks that the voice exists and speaks the chosen language, and tells you in plain words if it does not.

Speaking speed

Slower, normal or faster. This is a delivery setting, not a way to fit more words: the script's word limit (about 2.5 words per second, 75 words for a 30-second ad) is fixed at normal pace. Slower suits elegant or warm scripts and older audiences. Faster suits playful scripts for a young feed, but listen to the result against the captions; speech that outruns reading defeats the point of having captions at all. Normal is the right default and most ads should stay there.

Tone and voice are different things

Tone is a script setting: elegant, playful, direct or warm. It shapes the words. Voice shapes the sound. They have to agree. An elegant script in a bouncy voice sounds like a mismatch even when both are fine alone, and a playful script read by a measured voice loses its jokes. Decide the tone first, from the product and the audience, then pick a voice that could say those words. The UGC-style article has more on why this agreement is what makes a presenter feel like a person.

Background

Three options: the presenter's own background, a flat color, or an image you upload. The own background is the natural default and the most believable for feed ads. A color makes the ad look composed and works for elegant brands with a strong accent color. Your own image can place the presenter in your store or against your product's world, as long as it is not so busy that the product clip, when it appears, fights with it.

Consistency across versions

If you make several ads for one product, as you should when testing hooks, keep the presenter, voice and background constant and change the words. Then any difference in results comes from the script, not from a different face. Clipelle stores brand defaults in Settings, so the same presenter and style prefill every new ad form.

A quick decision path

For most shops the choices resolve in a few minutes in this order:

1. Audience: who is watching, in one sentence. Put it in the Direction field too.
2. Language and accent: the market's own.
3. Presenter: a portrait look your buyer would take a tip from. Filter by gender if that narrows it.
4. Quality: Standard to test, Premium for the ad that will run.
5. Voice: keep the preselected one unless energy or accent is off; preview two alternatives at most.
6. Speed: normal.
7. Background: own, unless your brand has a reason for a color.

The form shows the estimate in credits and dollars as you change the length and quality, and the project page shows the real amount afterwards. A first 10-second Standard ad is free after e-mail verification, which is a reasonable way to hear how a presenter and voice sound together before any of this costs anything.

Pick the person your customer would listen to, let that person speak at a normal pace in the customer's own accent, and spend on movement only when the face is the whole picture. That is most of the decision.