How to make AI ads

Generator vs timeline vs eonik

Three different jobs get sold as one. Pick the tool for the job you actually have.

Use a generator when you need shots. Use a timeline when you are finishing a one-off film. Use eonik when the job is ads: what is worth making next, then the cut. The same doors, named by tool, live on compare.

Decide before you generate

Do not open a generator first. Open what is already running. Yours and the category. And what you have already tested. Pick one thing to change: the opening, the proof, the offer, or the end card. Not all four.

In eonik that consideration lives in the app (the slate): what is worth making next, with receipts you can check. It is labor, not a prediction. It does not pick winners. You decide. Then you make the cut.

If you skip this step you generate into a void. You get a Drive folder of “maybe” clips and no test. Spy tools stop here. That is why they are not the product.

Get the raw clips

Write one body that carries the proof, the offer, and the CTA. Then write five to ten hook openings that can sit on that same body without reshooting it. Pain, curiosity, a number, a pattern interrupt, social proof. Different openings, same pitch.

Shoot the body on a phone if you can. Generate b-roll or extra hooks in Veo, Runway, or Kling if that is faster. eonik does not care which model you used. Bring the files.

Do not try to prompt a finished ad. Generators are bad at brand, captions, safe zones, and “now make me eight hook variants of the same cut.” That is assembly work. Treat generation like a camera: get usable shots, then edit.

Assemble on Mac

Import the footage, the AI clips, and the brand kit. Put a hook on the body. Type, color, logo, captions, end card. On every frame, not as a Figma comment you remember at export. Reframe 9:16, 1:1, and 16:9 from the same master. Watch it. Approve it. Export it.

That is the job CapCut makes you repeat by hand, once per size, once per hook. Download, import, caption, reframe, brand, export, again. That shuffle is why people think they need “an AI video editor.” They need the ad finished. eonik is not CapCut for marketing. It is the place the research becomes the cut.

AI clips on the timeline exists inside that gate if you need a shot fixed. It does not ship without you. If a frame is wrong, you kill it. Taste stays human.

In the Mac app

The loop above is the job. The clicks live in five how-tos for the Mac app. Make your first ad, or open all five. Brand kit, start from an ad you like, hook variants, and export. Not another magazine essay.

What usually breaks

• Generating more instead of assembling. More clips is not more tests. A test is an upload-ready variant with one isolated change.
• Brand rules living outside the cut. If logo safe zones and type live in a PDF, every export needs a cleanup pass. Put the kit on the timeline.
• Three timelines for three ratios. Meta, TikTok, and Shorts do not share safe zones. Recutting each size by hand is how a morning becomes a week.
• Changing two things at once. New hook and new offer in the same file teaches you nothing when it dies.
• Stopping at research. A swipe file is not an ad. CapCut never watches the category. Spy tools never finish the cut.

Upload is yours

Export H.264 MP4s. 9:16 first. Add 1:1 if you actually run feed. Launch them in Ads Manager with stop rules you wrote down before you spent. Hook rate is an operator ratio: Meta’s 3-second video plays ÷ impressions. Hold rate is ThruPlays (completion or 15 seconds) ÷ 3-second plays. Meta does not publish a universal “good” number. Compare to your own baseline. You pause. You keep. eonik never writes to the ad account and never tells you a variant will win.

A common pattern: the body still holds, the opening is dying. Lock the body. Swap only the first seconds. Assemble those variants in eonik. Sandbox them yourself. Read the opening. That is the whole loop. Decide, make, approve, read. Then do it again next week.

FAQ