How to Write Better Prompts for Afrigen's AI Refiner

The AI refiner is the single biggest lever you have over your final video, and most creators barely touch it. They type three words, hit generate, and then wonder why the result looks nothing like what was in their head. The refiner can only expand on what you give it. Feed it a vague idea and it fills the gaps with generic choices. Feed it a clear one and it amplifies your intent. This guide walks through how to write prompts that actually get you the video you imagined.

Start with the four anchors

Every strong prompt answers four questions before anything else: who is in the shot, where they are, what they are doing, and how it should feel. "A trader" is weak. "A young woman selling fresh tomatoes at a busy Balogun Market stall, calling out to customers, warm and energetic" gives the refiner four solid anchors to build on. You do not need to write a paragraph yourself — that is the refiner's job — but you must supply the raw direction.

Be specific about place

Generic locations produce generic footage. "A city street" could be anywhere on earth, so the model defaults to a bland, placeless look. Name the actual place: Lagos Island at rush hour, a quiet street in old Kano, a beach bar in Tarkwa Bay, the red earth of a village road in the southeast. Specific places carry specific light, architecture, and texture, and the refiner knows how to describe them.

Direct the camera, not just the scene

A scene description tells the model what exists. A camera direction tells it how to show that to a viewer, and that is where footage starts to feel professional. Add one camera instruction per prompt: a slow push-in on a face, a tracking shot following someone as they walk, a low angle looking up to make a subject feel powerful, a handheld feel for energy. One clear camera move beats five adjectives.

Control the mood with light and time

Lighting is the cheapest way to change how a clip feels. Golden hour reads warm and aspirational. Overcast feels honest and documentary. Neon night feels modern and urban. Harsh midday sun feels raw and real. Pick the time of day on purpose — "golden hour" and "blue hour just after sunset" will pull completely different emotions out of the same scene.

Pick one subject and commit

A common mistake is cramming a whole story into one clip: "a man wakes up, drives to work, meets his team, closes a deal." AI video models work in short single moments, not montages. If you ask for four things you will usually get a confused blur of all four. Choose the single strongest moment and describe it richly. If you need the full story, generate each beat as its own clip and stitch them later.

Use the refiner's output as a draft, not gospel

When the refiner expands your idea, read it before you generate. It is a draft you can edit. If it added "snow" to your Lagos scene or made your market quiet when you wanted it loud, fix that line and regenerate. The refiner is fast and free to run — treat it as a back-and-forth, not a vending machine.

A quick checklist before you hit generate

• Did I name a real, specific location?
• Is there exactly one main subject and one main action?
• Did I give one camera direction?
• Did I set a time of day or lighting mood?
• Did I describe the feeling I want in one or two words?

Run through those five points and your hit rate climbs immediately. Better prompts mean fewer wasted credits, fewer regenerations, and footage that actually looks like the idea you started with. The model is powerful — your job is simply to point it in the right direction.