From one product photo to a finished ad: what happens in between

You upload a product photo, type a few sentences, click Generate, and some minutes later a vertical ad is ready to download. Between those two moments a fixed sequence of steps runs. Knowing what they are makes you better at the input, because every choice on the form lands in one specific step. This article follows one ad through the whole chain as Clipelle runs it.

Step 0: what goes in

Two things start every ad.

A product photo. One is enough. The product clip later in the chain is generated from this single image, so it should show the product sharply and completely. Product photos that make a good product clip covers how to pick one.

A product name and two or three sentences. What the product is, who it is for, what changes for them. The script is written from these sentences, and nothing is added that is not in them, so a thin description gives a thin script.

If you would rather not type, URL to video reads a public product page for you: name, description and the large photos are filled in, and you check them and tick the photos you want before anything is copied. There is no guessing involved; it reads what the page says.

One rule applies to all free text: it is treated as information about the product, never as instructions. If you type "make it funny" in the description, the script will not become funny. The tone setting does that.

Step 1: the script is written

The first step produces the spoken script. It follows a fixed order: a hook, the problem, the product, the benefits, and a call to action as the last sentence. Every one of those is a spoken line. There is no text card for the hook and no end card for the call to action, so the words have to do the work on their own.

The script has a hard word limit of about 2.5 words per second of ad. At 30 seconds that is 75 words; at 10 seconds it is 25, which only leaves room for a hook, the product and the call to action. If a first draft comes out too long, it is shortened to fit. A spoken ad script in 75 words shows what fits in that budget and what does not.

Several form choices land here:

• Language sets the language of the script, the voice and the captions at once.
• Tone (elegant, playful, direct or warm) sets how the lines are phrased.
• Offer (up to 80 characters) is the only way a deal gets into the call to action. If you leave it empty, the call to action asks the viewer to look, not to use a code.
• Direction (up to 200 characters) lets you name the audience or the emphasis: "for runners who train before work", "lead with the warranty".
• Your own script, if you provide one, replaces the written script. The word limit still applies.

If you ticked Review script first, the chain pauses here. You see three hooks, pick one, edit the words, and approve. Only then does anything get made. Without the tick, the first hook is used and the next steps start immediately.

Steps 2 and 3 run side by side

Once the script is final, two things are produced at the same time.

The product clip

One short clip is generated from your photo with slow camera moves: a gentle push in, a drift across. The movement is deliberately calm. A product clip that zooms and spins pulls attention from the voice, and the voice carries the argument. The clip does not invent what the photo does not show; it moves the camera over what is there.

The presenter

An AI presenter speaks the script. This is the step where most of your form choices land:

• Presenter quality. Standard is recommended and costs one credit per second. Premium has more natural head and hand movement and the best lip-sync, and costs more per second.
• Presenter. A portrait look, chosen from the picker.
• Accent, voice and speed. The voice is matched to the language and the presenter; speed is slower, normal or faster.
• Background. The presenter's own, a flat colour, or an image you upload.

The presenter is not a decoration on top of the ad. The presenter video is the base of the whole ad, picture and voice. Everything else is laid on top of it.

Step 4: the final edit

With both pieces finished, the final edit assembles them:

1. The presenter video runs from the first second to the last and provides the audio.
2. The product clip is laid over the presenter at fixed moments in the timeline, with a 0.5-second fade in and out. The viewer sees the product while the presenter keeps talking about it.
3. Captions of what the presenter says run over the whole video. Your brand style decides how they look: the accent colour highlights the word being spoken, and you choose the caption look and the font.
4. If you uploaded a logo, it sits small in the top right corner.

That is the complete list of what appears on screen. No music, no text cards, no end card, no button. The only text in the video is the captions, and the only voice is the presenter's. This is a deliberate choice: one voice, one message, nothing competing for the viewer's eyes.

Step 5: the project page

While the chain runs, the project page shows each step with its status: Waiting, Processing, Completed or Failed. You do not have to watch it; you can close the tab and come back. A typical ad is ready in minutes.

If a step fails, the page tells you what went wrong in plain words and offers Retry. A retry resumes at the failed step, so a problem in the final edit does not redo the script or the presenter.

When the ad is complete, the result comes first: the player, the download, and the facts about the video. Next to the estimate you saw on the form, the page shows what the ad really cost. One credit is one second of Standard presenter video; the exact figure for your ad is on the page.

What this means for the input

Working backwards from the chain, here is what pays off on the form:

• Write the description for the script step. Problem, person, change. The script cannot say what you did not tell it.
• Pick the photo for the clip step. Sharp, complete, room around the product, even light.
• Put the deal in Offer, not in the description. Offer is the only route into the call to action.
• Use Direction for emphasis. One sentence about who the ad is for or what to lead with.
• Review the script on your first few ads. Three hooks, one choice, and you learn what your descriptions produce.
• Set brand style once in Settings. Accent colour, caption look, font and logo are then prefilled on every form.

An example, end to end

An invented product: a foldable laptop stand for people who work at a kitchen table.

Description: "A foldable aluminium stand that raises a laptop to eye height. For people who work from a kitchen table and end the day with a stiff neck. Folds flat, fits in a laptop sleeve."

Form choices: 30 seconds, Standard, English, direct tone, Offer "Free shipping this week", Direction "for remote workers without a desk".

The script step writes roughly 75 words that open on the stiff neck, name the stand, give the eye-height and fold-flat benefits, and close with the free shipping and where to find it. The clip step pushes slowly in on the photo of the stand. The presenter speaks the lines in a direct tone. The final edit lays the clip over the presenter at the fixed moments for a 30-second ad, runs the captions with the brand colour on the spoken word, and places the logo top right. The project page shows the real cost next to the estimate.

That is the whole distance between one photo and a finished ad. Make the photo and the sentences good, and the steps take care of the rest.