Use case · Advertising
Turn one product photo and one offer into a batch of ready-to-test vertical video ads. A strategist agent plans the test, and every variant comes out staged, animated, and voiced — what an agency bills per video, this graph produces by the batch at provider cost.
Since ad platforms took over targeting, creative volume is the lever that’s left: accounts win by testing many angles and refreshing them before they fatigue. That makes ad variants a recurring line item — bought monthly, per video. This workflow moves that line item onto your canvas.
How it works
Four inputs, one fan-out. The strategist plans once; everything after the list generator runs once per variant.
- Feed in the campaign. The product and offer, the audience, a variant count, and one clean product photo — the same shot anchors every variant. (e.g. “Aurora Trail smart fitness watch · 20% off launch week · weekend hikers · 6 variants”)
- Plan the test. A strategist agent studies the photo and brief, then writes a test matrix: buyer personas × message angles (pain point, bold claim, social proof, offer urgency), and picks the strongest pairings to test.
- Write the hooks. A list generator turns each matrix cell into one spoken hook — 18 words or fewer, plain language, said in the first three seconds.
- Stage every scene. For each hook, an image-to-image model places the product from your photo into that hook’s world: native UGC framing, natural light, 9:16, upper third kept clear for caption text.
- Animate and voice. An image-to-video model adds a subtle handheld push-in; a text-to-speech model reads the hook.
- Mix and ship. The voiceover lands on the clip, and the finished ads collect in a preview grid — one ready-to-upload vertical ad per matrix cell.
One photo in, a test cell out
Every variant isolates one persona × angle pairing, so when one ad wins in the ad account you know why it won — and the next run can breed from it. Paste last week’s winning hooks into the audience or offer input, raise the variant count, and the strategist plans the next generation around them.
Make it yours
- Swap any model. Language, image, video, and voice are each one node. Trade the video model up for hero variants, down for cheap broad tests.
- Retune the strategist. The persona × angle matrix is a prompt you can read and edit — push it toward founder-story ads, comparison ads, or a single angle across ten personas.
- Change the format. Aspect ratio, duration, and resolution live on the image and video nodes. Go 1:1 for feed placements or 16:9 for YouTube in one click.
- Localize the batch. Point the voiceover at another voice or language and re-run — the same matrix ships in every market.
Models in this workflow
The template ships with no models selected — pick one per role when you open it. Called with your own keys; the bill comes from the provider.
| Role | Node | Works well with |
|---|---|---|
| Test-matrix strategist | Agent | Any strong language model |
| Hook writer | List Generator | Any language model |
| Scene staging | Image To Image | An image-editing model that keeps the product intact (Nano Banana, FLUX) |
| Ad animation | Image To Video | Veo, Kling, Seedance |
| Voiceover | Text To Speech | OpenAI TTS, ElevenLabs |
See Models & Providers to set up keys.
Next steps
- Open the template: Templates → Ad Creative Factory in the dashboard
- Product Video Generator — one cinematic hero video instead of a test batch
- Movie Trailer Generator — the same fan-out pattern, pointed at film
- All use cases