Kling 4.0: video-to-video AI that keeps your real motion

A video to video AI workflow that keeps your real footage and turns sayings into clips

10 generations$2.51 total

Disclaimer: This is an independent creative test made with AI, for demonstration only. It isn't an ad for any brand shown or named, and isn't affiliated with, sponsored or endorsed by them. All trademarks belong to their owners

Kling 4.0: video-to-video AI that keeps your real motion

I took four sayings we all use without thinking, head in the clouds, the elephant in the room, when pigs fly, a fish out of water, and made each one literally happen to me, using my own footage instead of a brand new AI character.

I filmed myself for about 12 seconds total, just standing around in my hallway doing four tiny poses. This is a video-to-video thing: my real clip goes in, and the only part that changes is the one impossible thing, so it still looks like a normal phone video of me.

Here's my exact process.

1. Film four quick clips, one per saying

First I film four short clips, 3 to 5 seconds each, one simple action, decent light. The trick is to pick sayings you can act out with a plain pose. Looking up covers head in the clouds and when pigs fly, standing with some empty space beside you covers the elephant, and looking down at the floor covers a fish out of water. I shot all four back to back and lined them up into one clip in the Edits app, but you can keep them as four separate files, whatever's easier.

2. Cut the long clip back into its four scenes

If you stitched them into one like I did, split the clip back apart at each cut so every saying is its own little scene. The reason I bother: the next step drops one impossible thing onto a whole scene, so each saying has to be on its own.

3. Turn one frame of each scene into a still

Now I grab a single frame from each scene and add just the one impossible thing to it, keeping me, my clothes and my hallway exactly the same. I keep my instructions dead simple: "separate my head and put it in a big cloud high above me", "put a massive elephant sitting right beside me with an awkward look like it's staring at me", "two pigs with angel wings flying in a circle above me", "a huge red fish flopping on the floor at my feet". One thing per shot. If I change more than one thing, it starts to look like a generated scene instead of my actual hallway.

His head floating inside a white cloud high above his body in the hallway, made with GPT Image 2.6 Flare

Exact prompt

Create a new photorealistic image. Keep the first reference image exactly: same bright white fluffy cloud with his head floating in it near the ceiling, same headless body standing below, same real riad lobby, same framing and light. Change only one thing: add a light rain drizzle falling from the cloud. THE VIBE: Head in the clouds, now with a little personal rain shower drizzling down from the white cloud onto him. Photoreal, whimsical, deadpan. THE SUBJECT: The man from the reference, unchanged, same grey t-shirt, black shorts, slides, same standing stance and clean surreal neck cutoff, his head still floating in the white cloud up near the ceiling. THE KEY DETAILS: Thin streaks of light rain falling straight down from the underside of the white cloud, a gentle drizzle over his headless body and onto the tiles right below the cloud. A faint wet sheen and a few soft splashes on the floor directly under him, and the underside of the cloud a touch greyer where the rain forms. Light rain only, a drizzle, not a downpour. Everything else identical to the reference. THE LIGHT: Keep the lobby's warm indoor light; a subtle wet sheen on the floor under the rain catches the light. THE CAMERA: Authentic phone photo, deep focus, everything sharp, fine natural grain. THE COMPOSITION: Identical to the reference, headless body in the lower two-thirds, white cloud with his head high up, now with light rain between them. Vertical 9:16. References: The first reference is the finished white-cloud head-in-the-clouds scene, keep everything, only add a light rain drizzle falling from the cloud onto him.

A large elephant seated beside him in the hallway with an awkward stare, made with GPT Image 2.6 Flare

Exact prompt

Create a new photorealistic image. Keep the first reference image almost exactly: same man in profile, same grey t-shirt, black shorts and slides, same real riad lobby, same warm light, and the same massive elephant with its awkward seated pose and embarrassed wide side-eye. Change only one thing: move the elephant FORWARD so it sits right beside him, at the same distance from the camera, instead of set back behind him. THE VIBE: The literal elephant in the room, the huge elephant now sitting right next to him on the same plane, almost shoulder to shoulder, still staring at him awkwardly. Photoreal, deadpan. THE SUBJECT: The man exactly as in the reference, same profile stance, real face, grey t-shirt, black shorts, slides. The massive elephant from the reference, same awkward seated pose with front legs propped and splayed, trunk resting down, the same embarrassed wide side-eye locked on him, now positioned right beside him at the same depth, close to him, not behind him. THE SETTING: The same real riad lobby, carved ceiling, amber walls, wood elevator, carved pillars, tiled floor, unchanged. THE LIGHT: The same warm indoor light; the elephant grounded with a real soft contact shadow on the tiles right next to him. THE KEY DETAILS: The elephant huge and grey, sitting awkwardly beside him at the same distance from the camera, those self-conscious eyes still on him. He stands in his normal stance as if nothing is unusual. THE CAMERA: Authentic phone photo, deep focus, everything sharp, fine natural grain. THE COMPOSITION: The man and the giant seated elephant side by side at the same depth, the elephant filling much of the frame right beside him, eyes on him. Vertical 9:16. References: The first reference is the finished scene, keep the man, the room and the elephant's exact seated pose and awkward look, only move the elephant forward to sit beside him instead of behind him.

Two winged pigs hovering near the ceiling above him as he looks up, made with GPT Image 2.6 Flare

Exact prompt

Create a new photorealistic image. Keep the reference photo exactly: same camera angle, framing, crop, room, light, his body, clothes, face, hands-on-hips stance and upward gaze. Change only one thing: add exactly two flying pigs with large angelic wings circling in the air above him near the ceiling. THE VIBE: The literal "when pigs fly", deadpan and a little majestic: two winged pigs gliding in a slow circle high above him like birds circling, while he stands hands on hips looking up. Photoreal, like a real phone photo with one absurd detail. IDENTITY LOCK: The man from the first reference, preserved exactly: his real face, beard, hair, grey long-sleeve t-shirt, black shorts and slides, same stance, hands on hips, looking up. THE SUBJECT: Above him, near the carved ceiling, exactly TWO plump pink pigs rendered realistically, like real pigs, not goofy or cartoonish, each with a pair of LARGE, majestic, pure-white angelic feathered wings spread wide in a gliding pose. The two pigs are flying in a circular orbit, one behind the other, following the same circular path around an invisible center above his head, both heading in the SAME rotational direction (counter-clockwise as seen from below): the left pig is flying toward the right of the frame in profile-ish three-quarter view, banking gently into the turn, and the right pig is further along the circle flying toward the left, also banking into the same turn, nose leading along the curve, tails trailing, bodies tilted into the bank like birds circling in a thermal. Both clearly facing along the circular direction of travel, never toward each other or toward camera. THE SETTING: The same real riad lobby, carved plaster ceiling, warm amber walls, dark wood elevator doors, carved stone pillars and tiled floor, unchanged. THE LIGHT: Keep the lobby's warm indoor light exactly; the two winged pigs lit by the same light, a soft glow on their white wings, small soft shadows near the ceiling. THE KEY DETAILS: Exactly two realistic pink pigs, big graceful white angelic feathered wings outstretched, flying in a circle above him with a clear shared direction of travel and banked bodies. He gazes up, hands on hips, unbothered. Keep both pigs and wing tips fully inside the frame with a margin from the top edge. THE CAMERA: Authentic phone photo, deep focus, everything sharp, fine natural grain. THE COMPOSITION: The man in his reference position and scale in the lower half looking up, the two winged pigs in the upper third, fully in frame, arranged on a circular path. Vertical 9:16. References: The first reference is the man and his real lobby, preserve everything, only add exactly two realistic winged pigs circling in the air above him, both flying the same circular direction.

A huge bright red fish with gold fins on the floor at his feet, made with GPT Image 2.6 Flare

Exact prompt

Edit the reference photo. Keep it exactly: same camera angle, framing, crop, room, light, the man (his real face, beard, hair, clothes, slides, arms-crossed stance looking down), the tiled floor, the water droplets, and the fish's pose, position, bright red body, cream belly, big dopey eyes, gaping mouth, red pectoral fins, tail and long golden whiskers. Change only one thing: the fish's back fin along the top of its spine. THE VIBE: The same deadpan "fish out of water" photo, one ridiculous creature detail upgraded. Photoreal, like a real phone photo. IDENTITY LOCK: The man from the reference, preserved exactly, unchanged. THE SUBJECT: The same huge bright red fish flopping on the tiles at his feet. Its back (dorsal) fin along the top of its spine is now much LARGER and taller, a big flared, sail-like crest of fin rising high above the back and running along most of the body, in a rich metallic golden-yellow colour with fine ribbed fin rays and a glossy wet sheen, matching the gold of its whiskers. Everything else on the fish stays the same. THE SETTING: The same real riad lobby, unchanged. THE LIGHT: Keep the lobby's light exactly; the gold fin lit by the same warm light with soft highlights and a soft cast shadow on the tiles. THE KEY DETAILS: Large gold back fin, bright red body, long golden whiskers, fish fully in frame, nothing else changed. THE CAMERA: Authentic phone photo, deep focus, everything sharp, fine natural grain. THE COMPOSITION: Identical to the reference: the man on the left, the fish large in the lower middle and right of the frame. Vertical 9:16. References: The first reference is the whole scene, preserve everything, only enlarge the fish's back fin and make it gold.

4. Animate each scene so it keeps your real movement

This is the part that sells it, and it's the whole reason to go video to video instead of generating a fresh clip. I animate each still in a mode that keeps my real footage underneath: my movement, my timing, my camera. The AI only moves the thing it added. So the cloud rains on me, the elephant breathes beside me, the pigs circle and the fish flops, and I'm still just me, moving exactly like I did when I filmed it. I do this once per scene, so four short clips here.

Kling 4.0 $0.58

Exact prompt

Create a new video where a man's head sits inside a big white cloud above his body near the ceiling of a real lobby, the cloud swells and morphs into a much larger, darker-bellied storm cloud, and heavy rain pours down onto the floor around him, while his head stays fully visible in frame the whole time and keeps moving exactly like his real head in @video_1. The man's real motion, timing and the camera of @video_1 are kept exactly, including every head movement; the cloud, the separated head and the rain come from @image_1. Photoreal, handheld phone look, warm indoor light, comedic and surreal. [SCENE 1] The opening frame is the real take: his body stands in the lobby exactly as in his real take, and his head sits inside a white fluffy cumulus cloud floating above him, the whole head and face clearly visible near the top of the frame with a clear margin from the top edge, with a light rain falling from the cloud's underside. His head inside the cloud is alive and moves exactly like his real head in the original take: it starts glancing down and to the side, then turns, sweeps and looks around, tilting and rotating the same way and at the same moments as his real head, looking down and across, his eyes moving with it. It never freezes and never leaves the frame. As the take goes on, the cloud visibly grows and morphs: it billows outward to the sides and downward, puffing up wider and fuller, its edges rolling and swelling, spreading across the ceiling and widening to fill the top third of the frame, its underside turning a heavier soft grey. The cloud grows around his head, never above it: his head stays in the same place in the upper part of the frame, fully visible with the cloud wrapped around it, and the cloud's top edge stays above his head but is never pushed so high that it carries his head out of the frame. The rain grows with it: it starts light and quickly turns into a steady heavy downpour, thick streaks of falling water coming from the whole underside of the cloud, splashing hard on the tiles, the floor getting wet and shiny, droplets bouncing off his shoulders, shirt and shorts, his clothes getting wet and darker. His body keeps its real motion. His face is exactly his real face, no aging. Nothing else in the room changes.

Kling 4.0 $0.57

Exact prompt

Keep his real motion, timing and the camera of @video_1 exactly: the same room, light, framing, body, clothes and every movement of the take. Apply the look of @image_1: a massive elephant sits on the floor right beside him in a seated position, filling the space next to him, with an awkward, uncomfortable look in its eyes as it stares at him. The elephant is a real, heavy animal: it stays seated, its huge body barely moves, with slow breathing, a slow blink and its eyes tracking him as he moves and turns. It stays beside him and never moves behind him. His body, face and motion are exactly his real take, no aging. No other changes to the room.

Kling 4.0 $0.57

Exact prompt

Keep his real motion, timing and the camera of @video_1 exactly: the same room, light, framing, body, clothes and every movement of the take. Apply the look of @image_1: exactly two realistic pink pigs with large white angelic feathered wings fly in a slow circle near the ceiling above him, both heading in the same direction around the same circular path, banking gently into the turn like birds circling. Their wings beat with slow, powerful, graceful flaps and the pigs keep circling continuously for the whole take, never stopping, never facing the camera. He keeps looking up at them with his hands on his hips exactly as in his real take. His face is exactly his real face, no aging. No other changes to the room.

Kling 4.0 $0.59

Exact prompt

Create a new video where a huge bright red fish with long golden whiskers and a large gold back fin thrashes violently on the dry tiles of a real lobby at the feet of a man who stands deadpan with his arms crossed, looking down at it. The man's real motion, timing and the camera of @video_1 are kept exactly; the fish and the water around it come from @image_1. Photoreal, handheld phone look, warm indoor light, comedic and absurd. [SCENE 1] The opening frame is the real take: the man stands at the left with his arms crossed, looking down. The huge red fish lies on the tiles in front of him, mouth gaping. Straight away it goes wild: it slams its whole body against the floor and throws itself almost clear of the ground in big violent leaps, tail whipping high, the body bending hard into a curve, the head and chest lifting off the tiles for a split second before crashing back down. Every landing sends big sprays of water and splashing droplets flying across the tiles, wet slap marks and puddles spread around it, the golden whiskers whip through the air, the gold back fin flares open and snaps shut, and the mouth gapes wide with wide-eyed panic. The fish is heavy and wet like a real animal, rendered as a real tangible creature, and it keeps thrashing and half-leaping for the whole take without calming down. The fish stays in the lower right of the frame in front of him, never leaving the frame. The man never reacts beyond his real take, deadpan, arms crossed, face exactly his real face, no aging. Nothing else in the room changes.

5. Stitch the scenes back together

Last step, I put the four animated scenes back in a row into one video. Straight cuts, no transitions. If you want music or a caption, drop it in here or in your own editor.

Combine clips $0.00

Exact prompt

Combine the clips in order

Why it works

The whole thing lands because the footage is boring and real. It's just me in my hallway, filmed on a phone, and then one thing in the frame is impossible. If I'd generated a brand new, cooler version of me it would look fake and the joke would die. Keeping my own real footage underneath is the reason it still reads as me.

Your turn: pick one saying you can act out with a pose and film yourself doing it for 3 seconds. That's an AI video of yourself that still moves like you, with one impossible thing on top.

And if you'd rather not fiddle with the cutting and the frames, Emero does the splitting, the stills and the animating for you and keeps your clip's real motion, at cost.

Total for this workflow$2.51

Get one of these every day

I test new workflows daily and send them with the exact prompts, the models and the costs

6,526 generations tested so far

More workflows

See all