AI Filmmaking Program · Course 4
Seven lessons to turn loose ideas into a real production: build purposeful sequences, give your scenes physical movement, choose the right camera movement for each emotion, and bring it all together in a structured scene—ready for any AI video tool you have on hand.
Lessons
Stop generating unrelated images: build a four-scene sequence with a narrative arc before spending your first generation.
Break any movement down into force, sequence, camera, and reaction—and watch your video stop looking generic.
Choose between dolly, zoom, and dolly zoom, knowing exactly what feeling each one creates for the viewer.
Lock your character’s identity once and compare the same prompt in two different tools.
Learn the cuts that support any edit — and stop letting your video “jump” from one clip to the next.
Master the camera distances directors use to decide what each shot needs to show.
Structure an entire scene in time blocks and complete the full production workflow.
Course 4 · Lesson 1
By the end of this lesson, you’ll plan a four-scene sequence with a narrative arc—not four disconnected images competing for attention—ready to generate in any image tool you already use.
You already generated a good image and, on the next try, a second image of the same subject—and the two looked like photos taken on different days, with no connection between them. The problem isn’t bad luck: each image was created in isolation, without a plan behind it. A campaign only works when the scenes tell a story in sequence—and that’s decided before you generate the first image.
↓ role to study
Every cinematic image follows a silent rule: the viewer needs to understand the main subject in less than a second, without searching. If the scene has two or three potential protagonists—a product, a person, a background detail competing for attention—the eye gets stuck, and the image, no matter how beautiful, communicates nothing.
This is the most common mistake people make when they start generating images to promote their own business: requesting one image at a time without thinking about the whole set. The result is a collection of beautiful standalone photos that don't tell a story together — because they were never designed to from the start.
A neighborhood bakery owner launching a line of celebration cakes learned this firsthand: she generated six unrelated images for the launch, each with a different focus (the cake, the shop, the decorating hand, the counter), and the set looked like four different campaigns, not one.
This may look like a programmer’s flowchart, but it works like a conveyor belt in an industrial kitchen: each station does one job, and what comes out ready at one goes to the next. Workspaces designed for image sequences (Freepik is one of them) organize production into four stations: script (divide the story into scenes), character (create the protagonist’s or product’s visual identity just once), draft (a quick composition sketch, without worrying about detail or color) and finish (the final version, with real lighting and color).
The advantage of separating things into stages: you solve one problem at a time. Decide on the story without getting distracted by aesthetics; decide on the composition without getting distracted by color; only then move on to the final image—already knowing the foundation is right.
The bakery from the previous lesson used this workflow to launch its holiday line: a script with four scenes, the main cake described just once as a “character,” quick drafts of each scene, and only in the final step, the warm finish of the counter lighting.
Common mistake
Jumping straight to the final polish without going through the draft. The image looks beautiful, but the sequence makes no sense—each scene is reinvented because no step tested the composition before spending on the final generation. Test the composition cheaply (draft), spend big only afterward (finishing).
Four scenes are enough for a complete arc: general (establishes the setting), character (connects with a person or product), close-up action (builds tension or anticipation) and general again (wraps up, showing the result). It’s the same logic behind a 15-second commercial or an entire promotional album.
For the bakery: the display case early in the morning (wide shot, establishes the place) → the pastry chef decorating the first cake in the lineup (character, connects emotionally) → hands using a piping tip, in close-up (action, builds anticipation) → the party table ready, with the cake in the spotlight (wide shot again, closes the story).
When you reuse a result already created in the workflow (the script, the character, an earlier draft), some generators use a symbol in front of the name—a reference, usually marked with "@"—and that’s just a way of saying "use this existing thing," not syntax you need to memorize.
The process, in four steps
One point of honesty that can save you frustration: the first result is rarely perfect. Getting the composition wrong in the draft is cheap; that’s what it’s for. The ready-to-use prompt for the practice below is in English because most image generators understand technical terms better that way — you only need to replace the text marked between the signs < >.
Practice now 0/4 done
Leave with a list of four written scenes and at least two of them generated, keeping the subject the same—in ~12 minutes.
Image tools with a step-by-step workflow like the one in this lesson usually have a trial plan with a few generations per day, and that limit changes frequently. So write out the full list of four scenes before generating anything: test the sequence on paper instead of using one try after another on screen.
Cinematic scene sequence, 4 separate images, same subject and same visual style across all of them. 16:9 aspect ratio, consistent lighting and color palette throughout. SCENE 1 (establishing): Wide shot of <describe your place — shop front, kitchen, storefront> in the morning light. Calm atmosphere, no people close to camera. SCENE 2 (character): Medium shot of <describe your main person or product> doing the first step of the work. Natural expression, warm light. SCENE 3 (action, close-up): Close-up on the hands doing the key action of the work — the moment that matters most. Shallow depth of field. SCENE 4 (closing, wide): Wide shot of the finished result, same location as scene 1, now showing the outcome. Same lighting style as before. Keep the same subject, same clothing, same location style across all 4 images. Photorealistic, cinematic lighting, subtle film grain, no text, no watermark.
You have a sequence of four scenes with a beginning, middle, and ending — no more loose images competing for attention. It’s the same logic behind any video or visual campaign going forward.
Summary
Course 4 · Lesson 2
By the end of this lesson, you’ll write a movement prompt by breaking it down into force, sequence, camera, and human reaction—instead of using a vague adjective like “fast” or “dramatic”—and see the difference in the result.
Asking the AI to “make a dramatic movement” is like asking a butcher to “cut it nicely” without specifying the thickness: everyone understands something different. AI video tools understand physics — force, timing, weight — not adjectives describing a feeling. Without breaking down the movement, the result comes out generic or broken, no matter how many times you try again.
↓ role to study
Strength, timing, and weight: that’s all it takes to make any movement convincing. Strength is what starts the movement (an impact, acceleration, gravity). Time is how long each part of the action takes. Weight is how the object or person reacts physically — something light sways, something heavy resists.
A butcher at a gourmet meat shop understands this without ever hearing these terms: he knows that slicing a piece of picanha with a sharp knife has an initial force (the pressure of the blade), a duration (the cut isn't instant; it has a path), and weight (the meat gives and sways in a specific way, unlike a vegetable).
When you write “dramatic knife cut,” the AI doesn’t know how to translate “dramatic” into physics—it improvises. When you specify the force, timing, and weight, it has something to execute.
Motion generation tools like Seedance follow a sequence rule: the big happens first, the small next, the residue last. When slicing picanha: the blade goes in (big), the meat opens up (medium), and only then do the juices run and a wisp of steam rise from the hot plate (residue).
Ignoring this order—describing everything together, with no sequence—is the most common reason a movement comes out “wrong” even when the request is technically correct: the AI doesn’t know what comes before what, so it mixes everything together in the same moment.
Common mistake
Describe only the climax of the action, without a sequence or aftermath. "The knife cuts the meat" is just one instant — without the before (the pressure building up) or the after (the juices, the steam), AI delivers a dry, weightless cut that's hard to believe.
Beyond the scene’s physics, the movement needs to define the camera: where it is and how it moves (static, following closely, shaking slightly as if handheld). It also needs to define the human reaction—but show it through the body, never name it.
Instead of writing "he’s tense," describe what tension does to the body: fingers gripping the knife handle more tightly, breathing that changes pace, a gaze fixed on a single point. It’s the same principle a good actor uses: show, don’t announce.
After the cut, something keeps moving even though no new force is being applied—that’s residual movement: the juice still dripping, the steam still rising. It’s what makes the scene feel real, rather than like a photo that moved for a second.
The four layers of a motion prompt
Measured details help more than you might think: “a cut of about 2 seconds,” not “a quick cut.” Concrete numbers give the AI something to execute; vague adjectives leave the decision to chance.
Practice now 0/4 done
Walk away with a complete motion prompt (force + sequence + camera + reaction) for a scene from your business, and test it in a video tool—in ~12 minutes.
Motion generation tools usually offer a few free generations per day, and that limit changes frequently from one tool to another. Write the full request first, review the four layers on paper, and only use a try once you’ve revised it—that way, you make fewer mistakes on screen.
Physical motion sequence, realistic weight and timing, no exaggeration. FORCE: <descreva o que inicia o movimento — uma faca cortando, uma massa sendo sovada, uma porta se abrindo>, applied with natural, controlled pressure. SEQUENCE: main action happens first (about 1.5s), followed by a smaller secondary motion (about 1s), followed by a residual motion that lingers (steam, dust, liquid, fabric settling) for about 1s after. CAMERA: <descreva a posição — handheld close-up, static medium shot, slow tracking>, staying close to the action, natural slight movement. HUMAN REACTION: show tension or focus through the hands, breathing and gaze — do not name the emotion, only show the physical signs. Realistic physics, natural lighting, no slow-motion unless stated.
You wrote a motion prompt by breaking down the physics, not the feeling—and saw the result respond to what you asked for, not to chance.
Summary
Course 4 · Lesson 3
By the end of this lesson, you’ll choose among three camera movements—dolly, zoom, and dolly zoom—knowing exactly what emotion each one creates, and write the right prompt for a scene in your business.
"Move the camera closer" and "zoom in" seem like the same thing — but they aren't. Confusing the two is the most common reason a scene looks "flat" when you wanted intimacy, or "cramped" when you wanted distance. Each camera movement is a different gear: using the wrong one won't stall the car, but it won't take you where you wanted to go either.
↓ role to study
O dolly is the camera actually moving through space — as if you yourself took a step forward or backward. This creates real depth: the foreground, subject, and background move at different speeds, and your brain reads that as “I’m entering this place.”
O zoom is different: the camera stays still, and the lens changes focal length. The subject gets larger or smaller on screen, but the depth doesn’t change — the result feels flatter, like watching from a distance through binoculars without moving from your spot.
A flower shop owner feels this difference firsthand: a dolly in slow until a freshly delivered bouquet conveys intimacy, as if the customer were slowly moving closer; a dolly out revealing the entire store filled with flowers conveys scale — the size of the inventory, the abundance of the place.
O dolly zoom combines the two movements in opposite directions: the camera pulls back while the lens zooms in (or vice versa). The main subject stays the same size on screen, but everything around it distorts—the classic dizzying or sudden realization effect that cinema has used for decades.
For the florist: the moment a bride sees her bouquet for the first time and her expression changes—a subtle dolly zoom at that instant amplifies the feeling that “the ground moved,” without needing a caption that says “she was surprised.”
Common mistake
Asking for a “dramatic zoom” when the intention was intimacy. Zoom brings the subject closer, but it doesn’t change the sense of depth—the result feels observational, almost documentary, when the goal was to make the audience feel like they’re entering the scene. For intimacy, ask for a dolly; for observation, ask for a zoom.
One important point of honesty before you continue: tools like Kling read your prompt and try to fit it to a familiar movement — they don’t know on their own that you want “intimacy” or “surprise.” It’s up to you to name the right movement (dolly, zoom, dolly zoom), not just the emotion you want.
This may look like a decorated technical menu, but it works like the gear panel of a delivery vehicle: you don’t invent a new gear for every situation—you choose from the ones available, knowing what each one delivers. Along with the three you’ve already seen, it’s worth learning a few other common terms: tilt (the camera tilts up or down without moving, useful for revealing something from the bottom up), handheld (handheld camera, slight shake, a sense of realism and urgency) and crane (the camera rises and pulls back at the same time, creating grandeur and scale).
None of these replaces a dolly, zoom, or dolly zoom—they add to your vocabulary, each one suited to a specific situation.
Quick decision guide
It’s worth iterating: if the first result was close but not exactly what you wanted, reinforce just the movement word (“slower dolly in,” “more subtle dolly zoom”) instead of rewriting the entire prompt.
Test yourself
The flower shop wants a video that makes the customer feel like they’re “walking into” a store full of flowers, with real depth between the rows. Which movement works best?
Practice now 0/4 done
Walk away with three versions of the same scene from your business—one with a dolly move, one with a zoom, and one with a dolly zoom—and decide which best communicates the emotion you wanted—in ~15 minutes.
Testing all three moves uses three generations, and most AI camera tools have free plans with only a few attempts per day (the number changes frequently). If your quota is limited, test just two of the three moves today and save the third for tomorrow—nothing gets lost between sessions.
Cinematic camera movement, realistic physics, natural pacing, about 3 seconds. SUBJECT: <descreva o assunto — um buquê, um prato, um produto do seu negócio>, in a real environment with visible depth (foreground, subject, background). CAMERA MOVEMENT: <escolha um: "slow dolly in, camera physically moving closer" / "slow zoom in, lens only, camera stays static" / "dolly zoom, camera pulls back while lens zooms in, background distorts">. Natural lighting, no exaggerated speed, smooth and controlled movement.
You saw the same scene change its feel side by side, just by choosing a different movement — and now you choose based on the emotion you want to evoke, not trial and error.
Summary
Course 4 · Lesson 4
By the end of this lesson, you’ll create a locked character sheet—the same face, the same clothes, from any camera angle—and compare the same scene in two different tools, choosing the one that worked best.
An independent social media manager working with a gourmet butcher shop needs a recurring character — a “master grill chef” who appears in the videos every week — without hiring an illustrator or starting the design from scratch for every post. The problem isn’t a lack of talent: it’s not locking in the identity once, at the start.
↓ role to study
A character sheet — the same face, the same clothes, viewed from the front, side, and back, described just once — works like a shipping label: no matter which carrier delivers it, the contents of the box stay the same. Without that label, each tool (or each new prompt) “repackages” the character its own way, and the identity is lost.
The freelance social media manager who runs the gourmet butcher shop’s accounts created this profile just once for the “grill master”—specific apron, graying mustache, the way he holds the skewer—and started reusing it in every new piece instead of describing the character from scratch for each video.
Apply the same “locked character sheet” logic beyond a video character:
Apply the same logic
Newer tools like Luma were designed for natural conversation, not isolated technical commands. Instead of writing a new, complete request for every try, describe the scene like a director (“low camera, slowly rotating around the character”) and then refine what’s already there (“keep everything the same, just add a spark in the background”).
This changes the workflow: you build the scene in layers, one instruction at a time, instead of trying to get everything right all at once in one giant paragraph.
The advantage of speaking in layers only exists if you use the refine feature, not the restart feature. Treating each sentence as a separate command—ending the conversation and starting a new prompt for every adjustment—throws away the context built up to that point, and the character’s identity may change without anyone asking for it.
Common mistake
Treating each prompt as an isolated command. Ending and restarting with every adjustment (“now generate it again with the character smiling”) instead of refining the previous result (“keep everything the same, just change the expression to a smile”) makes the AI reinterpret the scene from scratch—and the outfit, pose, or setting can change without warning.
The right approach is always to start from what already exists: “keep everything the same, just...” — and only then describe the one change you want.
With the same character sheet in hand, it’s worth sending the same scene to two different tools—Luma and Runway are common examples—and comparing the results side by side. One may get the facial expression right; the other, the camera movement. Neither always wins: the right choice depends on the scene, not on loyalty to a single brand.
The butcher shop’s social media manager started testing simpler scenes (the character standing still and talking) in one tool, and scenes with more daring camera movement (orbiting the character) in the other—always saving the version that came closest to what she expected as that week’s standard.
The four steps
The AI guide starts from scratch with every new request, but you don’t have to: once written, the character sheet is the most reusable document in this entire lesson.
Practice now 0/4 done
Leave with a written spec sheet for a character (or mascot) for your business, tested in at least one AI video tool—in ~15 minutes.
Testing in two tools uses twice as many generations, and free AI video plans often limit how many attempts you get per day—the number varies widely across tools and changes frequently. If your quota only allows one tool today, test with that one and save the comparison for when you have access to the second; the written worksheet is already the main benefit of this exercise.
You have a reusable character sheet and have tested the workflow of refining instead of starting over — the foundation for any video series that needs the same look every week.
Summary
Course 4 · Lesson 5
By the end of this lesson, you’ll recognize and apply the four cuts that support most video editing — and know how to explain, for any of your clips, why a cut works or “jumps.”
The owner of a neighborhood bakery generated four beautiful clips for the launch of the holiday line—and when he joined them in a free editor, the video felt “choppy,” as if pieces were missing. The clips were good; the problem was where and how they were cut. Editing isn’t about pretty transitions—it’s about what carries over from one clip to the next.
↓ role to study
Editing isn’t about flashy transitions — it’s about the relationship between two clips. A cut works when something carries over from one side to the other: the movement continues, the composition stays similar, or the scene’s meaning moves forward. When nothing carries over, the cut “shows” — the audience feels the jump, even if they can’t explain why.
It's the same logic as a production line passing a box to another line: if both are at the same height and moving at the same speed at the moment of transfer, the box changes belts without anyone noticing the exact moment. If the height or speed doesn't match, the box shakes, almost falls — and everyone notices.
The bakery owner realized this when rewatching their video: the clips were good on their own, but each cut happened at a random moment, with nothing in common between the end of one clip and the beginning of the next.
O jump cut removes a piece of time: the camera stays in the same place, but the subject's position suddenly changes from one clip to the next — useful for showing a quick passage of time, such as the progress of a piece of handwork.
O match cut connects two clips through shape: a round object at the end of one clip cuts to another round object, the same size and in the same place on screen, at the start of the next. It's as if one "turned into" the other.
For the bakery: a round loaf centered in the frame cuts to a round celebration cake in the same size and position—the eye reads the change as a continuation, not an unrelated cut.
Before
Clips joined in the order they were generated, without thinking about composition or placement—the cut from the bread to the cake “jumps” because nothing repeats between the two frames.
After
The same pair of clips, reordered: the round, centered bread cuts to the round cake, also centered and the same size on screen—the sense of a jump disappears.
Result: zero new clips generated—only the cut point and sequence changed.
O cut on action ("cut on action") hides the cut in the middle of a movement already underway — the hand closing the order box, for example. Cutting at that exact moment, instead of during a pause, makes the eye follow the movement rather than notice the clip change.
As for the cross cutting ("parallel editing") alternates between two events happening at the same time in different places — the bakery kitchen finishing the last cakes, and the venue welcoming the first guests to the party. The back-and-forth builds anticipation: the audience feels the two storylines moving toward a meeting.
Test yourself
You’re editing two clips: a round, centered loaf of bread cuts directly to a round, centered birthday cake, also the same size on screen. What technique is this?
O double cut repeat a moment — show the same gesture twice in a row — from slightly different angles — to emphasize a moment worth highlighting, such as cutting the main cake at the party. The fast cutting strings together short, frequent cuts to create rhythm and intensity, useful at the climax of a video, but tiring if used all the time.
None of these techniques is “the right one”—each one creates a different mood. An editor’s job is to choose the technique based on the feeling the scene needs to create, not on which one is easiest to apply in the editor.
You don’t notice a good cut. You notice a bad one.
Practice now 0/4 done
Leave with two of your clips (generated or real) recut using at least one of this lesson’s techniques—in ~12 minutes.
You’re only rearranging files that already exist—no original clips are deleted, and undoing an edit costs nothing. If the result doesn’t convince you, go back and try a different cut point; mistakes here don’t use up any generation quota.
You edited two clips using a purposeful cutting technique—and trained your eye to notice, in any video, why a cut works or “jumps.”
Summary
Course 4 · Lesson 6
By the end of this lesson, you’ll choose the right camera distance for each moment in a scene—from the wide shot that shows the setting to the close-up that shows the feeling—and name the distances directors use.
A photographer helping a flower shop owner put together a photo catalog shot everything from the same distance — always full body, always 2 meters from the product. The catalog was technically correct and emotionally flat: not a single photo invites the customer to come closer, because none of them actually does.
↓ role to study
A wide shot answers “Where are we?”; a close-up answers “How does this feel?” The golden rule of cinema is simple: start wide to establish the world, move closer gradually to reveal emotion, and end on a close-up for maximum impact.
For the florist’s catalog, this means not using the same distance in every photo: a wide shot of the shop provides context (“this is where flowers are born for you”); a close-up of a petal with a drop of dew sparks a desire to buy in a way that no full-length photo can.
It’s not about which distance is "better" — it’s about the distance each moment in the story needs.
From widest to tightest: the wide shot (or establishing shot) shows the whole setting — the flower shop seen from outside, with the street around it. The wide shot moves in a little, showing the main space without focusing on anyone yet.
O American shot frames the person from the waist up, showing their body and action — the florist arranging a bouquet with both hands visible. The medium shot moves in a little closer, focusing on the main interaction. The medium close-up already makes facial expressions easy to read. And at the end of the scale, the extreme close-up shows just one detail — the dewdrop, the texture of a petal — that carries the emotional weight of the entire image.
Before
The entire catalog in medium shots, always at the same distance — technically correct, but none of the photos invites you to look closer or understand the place.
After
The same session, varying the distance in each photo: wide for context, medium for the florist’s action, extreme close-up for detail—each photo serves a different purpose in the catalog.
Result: zero new photos taken—the same session, only the framing choice changed.
If the location or action matters more at this point in the story, use a wide shot. If the emotion or reaction matters more, move in closer. If a single detail changes the meaning of the scene—the price on the tag, the ring hidden in the bouquet—use an extreme close-up, knowing it only works well when used sparingly, since it leaves out everything else in the frame.
The photographer who helps the flower shop puts it this way for the owner: “if you want the customer to understand the place, use a wide shot; if you want them to feel like buying, move in close.”
Test yourself
The flower shop wants to show a customer smelling a flower for the first time in a social media photo. Which shot distance best conveys this emotion?
An entire catalog (or video) of close-ups gets just as tiring as one made entirely of wide shots — varying the distance is what sustains the rhythm. The truth is simple: mastering distances isn’t about memorizing seven names; it’s about building the habit of asking, before each photo or scene, “what does this specific moment need to show?”
A wide shot shows the place. A close-up shows the feeling.
Practice now 0/4 done
Walk away with three versions of the same scene from your business—wide, American, and close-up—ready to compare side by side—in ~10 minutes.
Generating three versions of the same scene is no riskier than generating one: if a distance isn’t clear, just regenerate that specific version—the other two still work.
Cinematic photograph, same subject and same location, consistent lighting and color palette. SHOT: <escolha um: "wide establishing shot, showing the full place" / "medium wide shot, framing the person from the waist up, showing the action" / "extreme close-up on one small detail — a texture, a drop, a label">. SUBJECT: <descreva o assunto do seu negócio — o produto, a pessoa trabalhando, o local>. Photorealistic, natural lighting, shallow depth of field for close shots, no text, no watermark.
You have three distances for the same scene, each serving a purpose—the foundation for any catalog or sequence that needs to vary emotional pacing, not just angle.
Summary
Course 4 · Lesson 7
By the end of this lesson, you’ll structure an entire scene—from shots with a narrative arc to a technical script broken into time blocks—in the way AI video tools understand best, and know how to request the same scene again if it breaks halfway through.
You already combined a purposeful sequence, movement grounded in physics, an intentional camera, and a character with a locked identity. One last piece remains: organizing everything into a script the tool can follow from beginning to end without losing its way—and honestly recognizing that sometimes it will lose its way anyway, so the job is to adjust and try again.
watch this lesson on video (English · optional)
↓ role to study
You already saw in lesson 1 that a strong scene follows an arc: wide (space) → character (connection) → action (tension) → movement (energy) → wide again (closure). For a longer scene, that arc can expand to six shots while keeping the same logic.
What’s different in this lesson is how you deliver that arc to the tool: instead of a single paragraph trying to describe all six shots at once, you break it up by time blocks — a piece of the script for each time interval. Video tools don’t interpret it like a person reading a story; they follow the blocks in the order you arrange them, field by field.
For the launch of the bakery’s celebration line, this means that instead of writing “the bakery wakes up, the oven turns on, the cake is ready, the party happens” in a single paragraph, each moment becomes a block with its own timestamp.
This may look like a programmer’s form, but it’s a delivery route manifest: each stop has a time and a field for each decision—camera, action, light. You fill in the values; the punctuation around them is just a way to separate the fields, called YAML. You don’t need to memorize the punctuation—just fill it in.
00_02: camera: "wide aerial shot, dawn light" action: "forno da padaria acende, luz quente na janela" lighting: "amber, low warm light" 02_04: camera: "medium shot, static" action: "confeiteira decora o bolo principal, mãos em foco" lighting: "soft window light, warm" 04_06: camera: "close-up, slow dolly in" action: "bico de confeitar aplica o último detalhe" lighting: "warm, shallow depth of field"
Each time block has exactly one camera, one action, and one light — never all three mixed into the same sentence, which confuses the tool about what is camera movement and what is subject action.
Common mistake
Write everything in a single paragraph, without separating it into blocks. When camera, action, and light are mixed together in the same sentence, the tool can’t tell camera movement from the subject’s action—the result is unpredictable, even when the request is technically correct.
The character sheet from lesson 4 comes in here as a fixed reference, repeated in every time block — the same face, the same clothes, without redesigning them for each block. And between one block and the next, you name the transition, just as you named the camera movement in lesson 3: a simple cut for structure, one whip pan (quick camera spin) for speed, a smash cut (abrupt cut, with no smoothing) for impact, or a match on action (the video version of the cut on action from lesson 5) to keep the flow.
Naming the transition, rather than just describing the scene, is what makes the move from one block to the next come out as you planned — instead of letting the tool choose on its own.
Putting together the final scene
One final point of honesty, so you don’t get discouraged on your first try: even with a well-structured prompt, the movement may “drift” a little, the timing may not match exactly, and a transition may not come out as requested. That’s normal — the tool gets close to what you described; it doesn’t execute like a precision machine. Adjusting one block at a time and trying again is part of the process, not a sign that you did something wrong.
Practice now 0/4 done
Leave with a written script for a ~15-second scene for your business, divided into timed blocks with camera, action, and lighting—in ~15 minutes.
Writing the script in blocks doesn’t use up any generations — it’s paper work. If you have access to a video tool with a few free generations per day (the quota changes frequently from one tool to another), test only after the script is ready; it’s not your fault if the first generation comes out wrong — that’s how the tool works today, and adjusting is part of the process.
You structured an entire scene in the way video tools understand best—bringing sequence, movement, camera, and a locked character together in a single script. It’s the complete production workflow, from the first draft to the final scene.
Summary