01 Physics matters more than the effect

Anyone who’s ever asked an AI video tool for “smoke coming off a barbecue grill” or a “realistic splash of water” has seen the same result: something pretty, but cartoonish. The instinct is to blame the tool. That’s almost never the problem—the prompt described the subject, not the physics of the scene.

In major productions, effects that look real come from real-world physics: real fire smoke moves chaotically, and real water flows, splashes, and forms puddles in ways no one can program exactly the same twice. You don’t need to recreate this physically—you need to describe the behavior in words so the AI can simulate it.

A social media manager sees this every day: she generates a launch video for a new drink for a bar client, with ice splashing into the glass, and the splash looks like a bank commercial animation. She forgot to specify the weight of the ice, the direction it falls, and how the water breaks around it.

02 Chaos sells; perfection gives it away

Here’s a paradox that separates amateur prompts from professional ones: a perfectly symmetrical effect with even, predictable movement screams “computer” to the human eye. The real world is messy — and that messiness is what makes the brain accept an image as real.

Real fire never behaves the same way twice: the flame flickers, sends sparks in random directions, and changes height without warning. That’s why visual effects artists deliberately ask for chaos—and it’s exactly the opposite of a beginner’s instinct, which is to try to make everything “clean.”

The restaurant owner feels this pain all the time: he asks for “beautiful flames on the grill” and gets a cartoon flame, symmetrical, without a single stray spark. The dish beside it, with that fake-looking fire behind it, also loses credibility—the eye starts doubting the whole scene.

03 Light anchors the effect in the real world

In major productions, giant screens project the setting behind the actors—and the point isn’t just to show the background: light from the real screen hits the actors’ faces and clothes, creating real shadows and reflections. Without that light hitting correctly, the actor looks cut out and pasted into the scene, even with a beautiful background.

It works the same way with AI: the effect you ask for—a flame, a beam of light, a spark—needs to illuminate the rest of the scene in the same direction and color as the light that’s already there. If you don’t ask for this, the AI treats the effect like a sticker on top of the scene, with no connection to anything else.

The photographer sees this detail in candlelit wedding shoots: a lit candle beside the couple, without its warm light touching their faces, looks like an effect pasted in afterward. Explicitly asking for "the candlelight illuminates the couple’s faces, shadows flickering on the wall behind them" changes the entire scene.

04 The recipe: observe, describe, test

The first three readings from this lesson come together in a three-step recipe that works for any visual effect you request: observe (how that effect behaves in the real world — the weight, the chaos, the light), describe in separate fields (lighting, texture, motion, atmosphere, each one on its own, without mixing everything into a single sentence) and test (generate, compare, adjust one field at a time).

A social media manager uses this recipe to organize a request for sparks in a video for a woodworking shop: instead of writing “beautiful sparks flying,” she separates the light (warm, coming from the sander), the texture (small, dense sparks), the movement (a downward arc, pulled by gravity), and the atmosphere (fine dust suspended in the air). The result looks like an expensive corporate video.

You don’t need to memorize visual-effects terms—just these four fields, always kept separate. It’s the same logic as the production sheet you’ll see later in this program: one field per decision, so nothing gets lost in a run-on sentence.

01 What slow motion actually is

A slow motion — cinema’s “slow motion”—one of the audiovisual medium’s most powerful techniques: it stretches a moment and makes the audience feel that it matters more than the others.

In professional productions, it appears in moments that deserve weight: an important result, an emotional turning point, an impact. Outside those moments, it gets tiring—a whole video in slow motion is just a slow video, with no meaning at all.

The social media manager gets this request every week: "make a slow-motion grand opening video for the store, make it pretty." Without knowing where to use the effect or for how long, she tends to stretch it across the entire video — and the result is tiring instead of moving.

02 The secret is contrast: fast, slow, fast again

If everything is slow, nothing feels special. The impact of slow motion comes from contrast: normal pace before, time stretched at the moment that matters, then normal pace again afterward. That interruption is what tells the audience "pay attention here."

The social media manager applies this to the opening-day video: the normal movement of the street and people arriving, the ribbon-cutting stretched out in slow motion, and the applause returning to normal speed. Three blocks, one contrast—and the ribbon cutting becomes the center of attention.

03 Specify the frames per second, not just "slow motion"

"Slow motion" on its own is an incomplete instruction — AI needs to know how much to stretch. Always write the ratio: "120fps to 24fps" or, more simply, "5x slow motion". These expressions specify exactly how much to stretch time, as film professionals do.

It’s a good idea to change shots about every 2 seconds (wide, medium, close-up, a side angle) to create a cinematic rhythm — a scene that stays on the same shot the whole time gets tiring, even in slow motion. Describe small movement details too — blinking, breathing, a smile slowly appearing — so the character doesn’t look frozen.

The photographer uses this recipe to generate a video of a piece of jewelry slowly rotating for a catalog: tight shot, "5x slow motion," steady light, a glint that moves with the rotation. That detail—the same light from beginning to end—is what makes the rotation look truly filmed, not just stretched out.

04 Common mistake: asking for “slow motion” and stopping there

The restaurant owner fell into this trap when he asked for “a dish being flambéed in slow motion”: the flame rose slowly throughout the entire video, with no moment of normal pacing before or after—it got tiresome instead of impressive. Rewriting it with a fast-slow-fast structure and “5x slow motion” only for the moment the flame rises gave the same dish the weight it was missing.

01 The real world is never still

Call it particle systems any cloud of small elements moving together in a scene: smoke, dust suspended in the air, sparks, confetti, mist. They exist because the real world never stays perfectly still — and a scene without any of these elements feels sterile, digital, dead.

In an AI-generated scene, this absence is even more noticeable: with nothing moving through the air, the background looks like an empty studio set, not a real place. Particles solve this with little descriptive effort — and they solve another problem you’ll see in the next step.

The social media manager feels this when generating a video for a store’s anniversary — five years in business. A "clean" video, with nothing in the air, looks like an ad for an empty real estate property. With confetti and streamers in motion, the same setting becomes a real celebration.

02 The particle is a curtain, not a decoration

Here’s a professional technique few people know: particles hide the errors AI still makes. AI-generated video jitters during transformations — a shape changes in a strange way, an edge flickers or “melts” for a fraction of a second. Instead of trying to create a mathematically perfect transformation, professionals use physics to fool the eye.

Dense smoke hides the exact instant a shape changes. Haze softens shaky edges. Dissolving particles break up a predictable outline, making any imperfection look intentional. The viewer’s eye follows the smoke’s movement—not the small mistake behind it.

The restaurant owner uses this trick without knowing its name: when generating the transition from a raw dish to the finished dish, he passes a cloud of stove smoke through the middle of the change. The audience sees the smoke rising—not the moment the scene “jumped” from one state to another.

03 Where atmosphere comes in: light, density, wind

Three details determine whether the particle looks real or just like an effect floating over the scene. The light: good particles catch and scatter the light already in the scene — dust glowing in a beam of sunlight, smoke lit from behind. The density: how much you can see through it — thin lets the background show through, dense hides it on purpose. And the wind: the direction in which everything moves, giving the scene a coherent current of air, not particles floating aimlessly.

Combining more than one layer—background smoke with sparks in the foreground, for example—creates depth: each layer sits at a different distance from the camera, and the eye reads that as real space, not an effect pasted onto a flat image.

The photographer applies this idea to a studio still: a fine mist, side-lit, suspended between the background and the person being photographed. The result has a depth that a “clean” photo in the same studio would never have.

04 The recipe: describe the particle, the light, and the wind

Every particle prompt that works names three things: the type and density of the particle, the light it reflects, and the direction of the wind moving it. Without those three details, AI improvises—and the result usually comes out generic or, worse, with no visible effect at all.

01 Action is more than movement

A scene with lots of movement and no tension is just commotion — the eye gets tired and forgets it in seconds. Real action, the kind that stays in the audience’s mind, has four parts, always in this order: visual risk (something seems to be at stake), movement (the energy of the scene), expectation (the audience doesn’t know how it ends) and turn (the resolution, the payoff for that tension).

The question every good action scene leaves hanging is simple: “and now, how will this end?” Without that question, no matter how many effects you pile on, the scene doesn’t stick.

The social media manager feels this when generating a video for a café client: a waiter crossing the packed dining room, dodging chairs, to deliver the order before the birthday cake candle goes out on its own. Movement alone would just be a waiter walking fast — the nearly extinguished candle is what creates the question.

02 Tension comes from almost

The technique most often used in commercials and high-impact music videos isn’t shock—it’s almost. Something almost goes wrong, then works out at the last second. That "almost" is what makes the audience hold their breath for half a second, without any genuinely scary effect.

It even works with objects: a stack of delivery boxes slowly tips, one slips, and the delivery person catches the last package inches from the floor. No one gets hurt, nothing actually breaks—but the audience feels the entire “near miss.”

The restaurant owner applied exactly this in a behind-the-scenes video: boxes of ingredients stacked in the kitchen wobble with the movement in the dining room, one nearly falls, and the cook catches it without putting down the knife. Visual risk, with no real danger—just the feeling of “almost.”

03 One strong idea per scene

Five seconds can’t fit five ideas. The golden rule for anyone writing action for commercials and music videos is simple: one strong visual idea per scene. Trying to fit a “near fall,” a reveal, and a chase into the same five-second clip makes everything confusing—none of the three ideas has room to breathe.

The photographer applied this rule to a wedding video: the bride’s bouquet slips from her hand right at the edge of a puddle and is caught midair by a guest, inches from the water. One idea—the bouquet’s “near miss”—carries the entire scene. Nothing else competes with it.

The same goes for reveal scenes: a box opening, a product appearing, a curtain slowly opening—these are also strong visual ideas that don’t require any physical risk. The key is to choose just one idea and let it carry the full five seconds on its own.

04 The question that tests whether it worked

After generating any action scene, ask this simple question: would someone watching think, “He barely made it”? If the answer is yes, the formula worked—all four are there: visual risk, movement, anticipation, and a turning point. If the answer is no, one of those four is usually missing, most often the anticipation.

A social media manager uses this question to review her own delivery clip before posting: the waiter running looks great, but no one would root for him if there weren’t a candle almost going out in the background. Adding this simple detail changes the answer from “no” to “yes.”

01 What you feel is different from what you say

In film, people almost never say exactly how they feel. A powerful dialogue scene doesn’t come from the words—it comes from the emotion underneath them, the body language, the pauses, and the conflict no one names out loud. This has a name: subtext — the real feeling running beneath the dialogue.

When you ask the AI for dialogue and let the character “explain everything they feel,” the scene loses all its tension. Audiences like reading between the lines—taking that away from them takes away their reason to watch.

The photographer sees this in her own studio: a client who can’t decide which photo package to choose looks at the catalog in silence, smiles uncertainly, and says only "I’ll think about it." They never say "I’m afraid of spending too much"—but everyone in the room knows exactly what’s going on.

02 Reaction, silence, and conflict do the heavy lifting

Three tools carry a dialogue scene more than any line of dialogue: the reaction of the listener (sometimes the listener's face says more than the speaker's words), the silence (what’s left unsaid carries more weight than any dialogue) and the conflict — two people wanting different things in the same scene, without needing a shouting match.

When writing the prompt, describe the listener’s reaction as carefully as you describe the dialogue: a glance away, a pause before answering, a hand gripping the cup a little tighter. Those are the details the AI needs—not the emotion named outright.

The restaurant owner captures this kind of scene every day without realizing it: an unhappy customer never complains out loud, but the waiter notices how they fold the napkin, how long they wait before tasting a second forkful. A corporate video about customer service can use exactly this tense silence, with no complaint spoken.

03 The rule of thumb: if everything is said, the scene dies

Use this simple guideline when reviewing any dialogue prompt: if a character explains their feelings out loud, rewrite it by replacing the line with a physical reaction. The rule applies to any scene, commercial or personal — it’s the difference between a scene you feel and a scene that only gives you information.

01 Movement without a reason is not performance

Real people never move without a reason: fear creates hesitation, confidence creates direction, curiosity creates observation. Every gesture begins with an emotional motive before it becomes movement—that’s the foundation of AI character performance.

When the request describes only the action—“character walks to the door,” “character smiles”—the AI delivers exactly that, with nothing behind it. The result is technically correct and emotionally empty: a puppet walking, not a character feeling something. You need to write the reason behind the gesture.

The photographer feels this difference while generating an emotional couple’s portrait video—the moment the bride turns to see the groom for the first time on their wedding day. “She turns and smiles” gives you a generic, ad-like smile. “She turns hesitantly, still unable to believe it, and the smile comes slowly, as if she were taking in the scene” changes the entire character.

02 Describe the behavior, not the adjective

Small details carry big emotions: a gaze that drifts out of focus, a blink slower than usual, a tense jaw, a change in breathing rhythm. These signs say more than any words—and they’re exactly what AI needs to receive instead of a named emotion.

Asking for a “sad character” is vague: the AI improvises a generic, postcard-like sadness. Asking for the physical behavior of sadness — the averted gaze, the slow blink, the tense jaw — gives the AI something concrete to simulate, the same way a good actor builds a scene through the body, not an emotion label.

The restaurant owner applied this change in a corporate video about the kitchen’s behind-the-scenes: “tired cook after the rush” produced a face with decorative-looking fatigue. Rewriting it as “shoulders drooping, slower breathing, eyes fixed on the stove without really seeing anything” made the exhaustion look real.

03 The pause proves it’s alive

Real emotion takes time. A strong performance doesn’t react instantly — it processes before reacting, and that moment of silence is what creates realism. Body posture does the rest of the work: slumped shoulders, a raised chin, a heavier step reveal the character’s inner state before any bigger gesture.

In the prompt, this becomes a simple instruction: describe the pause before the reaction, not just the reaction. "Character reacts to the news" asks for an immediate response; "character stands still for a moment, taking in the news, and only then their shoulders slump" gives the audience time to feel the scene alongside the character.

The photographer uses this technique in a family shoot: the moment a father sees his newborn for the first time, in a pregnancy announcement video. Instead of an instant smile, she asks for "he freezes for a second, looking at the baby without moving, and only then his face opens into a slow smile"—the pause is what makes the scene moving.

04 The recipe: motivation, detail, pause

The three readings from this lesson come together in a three-field recipe, always in this order: subject (the emotion driving the gesture), detail (the physical behavior that proves this emotion — gaze, breathing, tension) and pause (the time the character takes to process before reacting). Together, the three are what separate an acting character from a puppet that only moves.

The photographer checks every character clip against this recipe before approving it: without a reason, the gesture seems random; without detail, the emotion feels generic; without a pause, everything seems too automatic. Fixing one field at a time usually solves the problem without having to generate everything again from scratch.

01 Shot distance is emotional distance

The distance between the camera and the character isn’t just an aesthetic choice — it’s the emotional distance the audience feels. A wide shot, with extra space around the character, creates a sense of detachment, even if the scene is warm. A emotional close-up does the opposite: brings the audience closer to a point where small details become impossible to ignore.

That's why the same scene, filmed from two different distances, tells two different emotional stories — even with the same dialogue, the same character, and the same light. The distance within the frame becomes the distance the audience feels from the character.

The photographer tested this while generating a video of a couple’s marriage proposal: in the wide shot, the two appear small in the middle of the garden—beautiful, but distant. In the close-up of their hands, at the moment the ring slides onto the finger, that same moment becomes something the audience feels, not just watches.

02 Empty space around a character also conveys emotion

The empty space around a character also conveys emotion. A character surrounded by open space, with nothing filling the frame, conveys isolation—even without any dialogue explaining it. It’s the opposite of a close-up: instead of moving closer, the scene deliberately pulls away, and that distance becomes part of the feeling.

In the video prompt, describe how much empty space there is around the character and where they are positioned in the frame — small and pushed into a corner creates more loneliness than centered; a large, empty room in the background carries more weight than a nearby wall.

The restaurant owner used this in a corporate video about the end of a busy day: alone, counting the cash after closing, small in the middle of the empty dining room, with chairs still out of place on the tables. The same moment, filmed up close and centered, would have looked like just “counting money”—from a distance and small in frame, it became real exhaustion.

03 Common mistake: moving closer or farther away without a reason

The photographer made this mistake during a family shoot with a teenager who wasn’t comfortable in front of the camera: she asked for a dramatic close-up right at the start of the scene, and the result felt invasive, almost embarrassing. Starting again with a wider shot and moving in gradually as he relaxed made the same scene feel genuine.

04 Move in with purpose: a slow push-in, over-the-shoulder, point of view

Three camera moves bring you closer on purpose, each for a different reason. A slow camera push builds tension gradually — the audience doesn’t consciously notice the movement, but feels the emotion growing along with it. The over-the-shoulder shot creates connection when there’s tension between two people in the scene. And the point of view — the camera where the character’s eyes would be — is reserved for the moment when the audience should feel exactly what the character feels, not just watch from the outside.

The practical rule: choose the movement based on the emotion already present in the scene, never for the effect alone. A slow push-in with no prior tension gets tiring; an over-the-shoulder shot with no conflict between the two characters creates no connection at all.

The photographer applies this to a family shoot: a grandmother reuniting with her grandson after months apart. The camera moves in slowly as she recognizes his face, building emotion little by little—it wouldn’t be the same scene with a direct cut to the close-up.

05 The recipe: choose by emotion, not by style

The photographer tests this recipe by generating the same scene twice—once following the distance the emotion calls for, and once with the generic distance she would use out of habit. The difference usually shows up in the first few seconds: the right version draws you in; the generic one just shows you the scene.

01 Every dialogue scene is an emotional negotiation

Two people talk, one asks a question, the other answers, everything seems logical — and yet the scene still feels dead. That’s because movie dialogue isn’t born from words alone: it’s born from conflict. One dialogue scene structure is what organizes that tension from beginning to end, instead of letting the conversation drift.

You already saw in lesson 5 that no one says exactly what they feel—that’s the subtext. Today you’ll learn the larger structure that organizes this subtext across an entire scene: the two people in the conversation need to want different things, even if neither says it out loud—that clash of desires is what sustains the scene.

The photographer applies this to an anniversary video: a seventy-year-old couple deciding whether to sell the house where they raised their children. She wants to keep the house for the memories; he wants to be free of the burden of maintaining it. Neither of them says it in those words—but the conflict carries the entire scene.

02 Silence and power determine who controls the scene

Emotion needs time to land—the silence before a reply creates tension that no line of dialogue can create on its own. And whoever controls the pace of that silence usually controls the scene: someone who answers right away seems anxious or weak; someone who waits before answering seems in control.

In the video prompt, this becomes a concrete instruction: describe not only who speaks, but who waits — and for how long. "Both respond right away" makes the scene balanced, with no hierarchy; "one responds right away, the other lets the silence stretch before speaking" tells the audience who is in control of that moment.

The restaurant owner uses this technique in a video about a negotiation with the supplier of a delayed ingredient: the supplier explains quickly, nearly tripping over their words; the restaurant owner lets the silence stretch before replying. His silence carries more weight than any words could.

03 The goal changes something, or the scene isn’t a scene

Every strong dialogue scene changes something emotionally by the end—someone finds courage, gives something up, or admits what they’ve been hiding. If nothing changes from beginning to end, what you have isn’t a scene: it’s just two people trading lines.

The photographer applies this to the scene with the seventy-year-old couple: she starts out resisting the idea of selling the house; when she hears her husband admit that he also feels the burden of maintaining it alone, she gives a little—not agreeing to sell, but stopping short of saying “never.” That small shift is the scene’s turning point.

04 The camera changes what the dialogue means

The same line, filmed in different ways, becomes emotionally different scenes—you saw this in lesson 7, about camera distance. In a dialogue scene, this technique has an added use: a close-up of the person listening while the other speaks makes the audience feel the reaction more than the line itself.

A wide shot with both characters divides the audience’s attention equally between them — good for showing the negotiation as a whole. Alternating close-ups of the two, especially of the person who’s quiet, reveal what the dialogue alone conceals.

The photographer uses this alternation in the couple’s scene: while he talks about the burden of keeping the household going, the camera stays on her face, silent, taking it in—it’s her silence, not his words, that carries the emotional weight of that moment.

05 The complete recipe: put together everything you’ve learned

The seven lessons in this course come together in one recipe for any character scene you generate from now on: write the subject behind the movement, the detail physical proof of the emotion, the pause before the reaction, the distance of camera that the emotion calls for and, when there's dialogue, the conflict between what each character wants, the silence that decides who's in control and the turn that changes something by the end.

Ricardo, who had owned a neighborhood restaurant for twenty years, wanted a promotional video to celebrate the anniversary: a short scene between him and his daughter, who wants to modernize the family menu. He wrote the conflict (he wants to preserve his father’s recipes; she wants to update the menu), left a silence before his response—he looks toward the kitchen before speaking—and ended with a small twist: he doesn’t agree with everything, but he’s willing to try one of her new dishes. The camera stays on his face during the silence, not on her as she speaks. The result doesn’t feel like a restaurant ad—it feels like a real family conversation.