The 5-Part Prompt Formula: Subject + Action + Camera + Light + Style

WHY ONE FORMULA WORKS FOR EVERY SCENE
Every AI video generator faces the same problem: it must turn words into moving pictures, and it can only animate what you actually describe. A prompt that misses a part forces the tool to guess, and guessing is where the fakeness comes from — a wandering camera, flat light, or an animal that stands still instead of moving.
The 5-part formula removes the guessing. Answer five questions, in this order, and you have directed the whole clip: who or what is on screen, what it does, where the camera is, how the scene is lit, and what it should look like. It works for a wildlife clip, a cooking video, and a comedy skit, because every moving scene has those five ingredients — only the details change.
Table of Contents
- WHY ONE FORMULA WORKS FOR EVERY SCENE
- PART 1: SUBJECT — BE SPECIFIC, NOT FANCY
- PART 2: ACTION — ONE CONTINUOUS MOTION
- PART 3: CAMERA — ONE SHOT, ONE MOVE
- PART 4: LIGHT — NAME THE SOURCE AND DIRECTION
- PART 5: STYLE — THE FINAL FILTER
- THE FILL-IN-THE-BLANK TEMPLATE
- THREE WORKED EXAMPLES, THREE NICHES
- WHAT GOES WRONG IN EACH PART
- FREQUENTLY ASKED QUESTIONS
- FINAL THOUGHTS
PART 1: SUBJECT — BE SPECIFIC, NOT FANCY
The subject is the anchor of the clip. Name it exactly, with one or two defining details that separate it from every generic version of the same thing.
Weak: “a bird.”
Strong: “a baya weaver bird with a half-finished hanging nest.”
Weak: “a man cooking.”
Strong: “a street vendor in a white apron flipping flatbread over an open flame.”
Two details are enough. Five details crowd the prompt and fight with the action for attention. Pick the details the camera will actually see: species, colour, clothing, age, or a distinctive object. Skip backstory — the generator cannot film “a bird who has travelled far.”
PART 2: ACTION — ONE CONTINUOUS MOTION
Action is where most prompts die. They describe a pose, not a movement, and the result is a still image with a slight wobble.
Use one continuous motion with a beginning and an end. “The weaver bird weaves a grass strand into the nest entrance, tugs it twice, then flies off for another strand” is a complete action. “A bird near a nest” is not an action at all.
Strong verbs do the work: swooping, pouring, unfolding, stacking, slipping, chasing. Static words do not: standing, sitting, beautiful, happy. If your prompt has the word “and then” more than once, you have packed two scenes into one clip — cut the second one and save it for the next video.
PART 3: CAMERA — ONE SHOT, ONE MOVE
Real footage always has a point of view. State the shot first, then add at most one camera movement.
Close-up: emotion and detail — an eye, a hand, a texture. Close-ups hide generator weaknesses because less of the scene is visible.
Wide shot: place and scale — a market at dawn, a riverbank before the action starts.
Slow push-in: gentle movement toward the subject. The most cinematic move for a reveal.
Tracking shot: the camera moves alongside a moving subject, keeping pace.
Low angle: shooting from below makes a subject look powerful and large.
One move per clip is the rule. A push-in plus a pan plus a zoom in ten seconds usually turns into a shaky mess, and the subject itself stops moving clearly because the generator is busy moving the camera.
PART 4: LIGHT — NAME THE SOURCE AND DIRECTION
Light decides mood and realism more than almost any other word in the prompt. “Nice lighting” gives the generator nothing. A named source and direction gives it shadows, contrast, and colour temperature to render.
Compare: “good lighting” versus “late-afternoon sunlight from the left, long shadows across the courtyard.” The second version tells the tool where the bright side is, where the shadows fall, and what colour the scene should feel like.
Use light to match the mood, not fight it. Golden hour suits warmth and nostalgia. Harsh midday sun suits a busy street scene. Soft window light suits an indoor moment. And add one imperfection — wet fur, cracked earth, a chipped mug — because perfect things look generated, while one flaw tells the viewer’s brain the scene exists.
PART 5: STYLE — THE FINAL FILTER
Style goes last because it filters everything above it. It covers the look, the aspect ratio, and the duration — the three things that decide whether the clip fits the platform it is made for.
End every prompt with a style line like: “Photorealistic, cinematic, 9:16 vertical, 10 seconds.” For a different project, swap the filter, not the scene: “3D clay animation style, warm minimal colours, 9:16 vertical, 10 seconds” turns the same action into a completely different video.
Never leave duration and aspect ratio out. Without them, you get whatever default the tool prefers, which is rarely what Reels, Shorts, or YouTube needs.
THE FILL-IN-THE-BLANK TEMPLATE
Copy this, fill one line at a time, and you have a complete prompt:
[Subject with 1–2 defining details] [one continuous action, beginning to end]. [Camera shot + one movement]. [Light source + direction + one shadow or colour detail]. [Style], [aspect ratio], [duration].
A filled example:
A baya weaver bird with a half-finished hanging nest weaves a fresh grass strand into the nest entrance, tugs it twice, then flies off for another. Close-up, camera steady on the nest. Morning sunlight from the right, warm gold on the grass. Photorealistic, documentary-style, 9:16 vertical, 10 seconds.
Five parts, five sentences or fewer, nothing extra. If a word does not change what the viewer sees, delete it.
THREE WORKED EXAMPLES, THREE NICHES
Wildlife. “A golden eagle with a scarred left wing swoops toward a mountain river, wings tucked, then flares open inches above the water. Slow tracking shot from the side, keeping pace with the bird. Golden hour sunlight from the left, long shadows across the rocks. Photorealistic, cinematic, 9:16 vertical, 10 seconds.”
Cooking. “A street vendor in a white apron flips flatbread over an open flame, catches it, and slides it onto a stack. Close-up on his hands, camera steady. Warm firelight from below, orange glow on his face. Photorealistic, documentary-style, 9:16 vertical, 10 seconds.”
Comedy. “A husband in a suit walks in carrying twelve shopping bags, trips on the doorstep, and catches himself just before falling. Wide shot from inside the living room, camera still. Bright afternoon light through the window, clean and cheerful. Cartoon style, 9:16 vertical, 10 seconds.”
Same formula, three completely different videos. Change the details, never the structure.
WHAT GOES WRONG IN EACH PART
Subject too vague. “A dog playing” gives the generator nothing. Fix it with species, place, and one detail.
Action too crowded. Two actions plus a twist ending in ten seconds becomes a blur. Keep one action; move the rest to the next clip.
Camera doing too much. Two moves in one clip fight the subject’s motion. Pick one move or none.
Light described as a feeling. “Beautiful light” is not a direction. Name the source, the direction, and one shadow.
Style missing. No aspect ratio or duration means the clip may not fit the platform at all. Always state both.
When a video comes out wrong, check the parts in order. Nine times out of ten, exactly one part is weak, and fixing that one part fixes the clip.
FREQUENTLY ASKED QUESTIONS
Do I have to use the parts in this exact order?
Yes, keep this order. Subject first anchors the scene, action second drives the motion, and style last filters the result. Reordering them does not break the generator, but it makes prompts harder to check — and checking is how you find the weak part.
How long should the finished prompt be?
Two to four sentences, roughly 40 to 80 words. One sentence is usually too vague, and a full paragraph usually contains contradictions.
Can I add dialogue or text to the prompt?
Only if the tool you use supports it, and keep it to one short line. Long dialogue in a ten-second clip crowds out the action, and on-screen text should be stated explicitly or the generator may invent garbled words.
Does the formula work for every AI video tool?
The structure works everywhere, but each tool interprets style words differently. Keep a small library of prompts that worked in your tool, grouped by niche, and reuse the structure with new details.
FINAL THOUGHTS
The formula is not a trick. It is directing, written down: a specific subject, one continuous action, one camera move, named light, and a style filter that fits the platform. Write prompts that way, and your clips will look deliberate instead of accidental.
Try it on your next video, starting from the template above. When you want ready-made starting points built on the same structure, browse the free prompts on PromptVilla — and create a free account to save the ones that work for you.
New to this series? Start with the first guide: How to Write AI Video Prompts That Actually Look Real.
