Skip to content
Skip to the prompts

Make a movielike this

Found footage · Director Mode · Seedance 2.5

Running from the Ophanim,
on one shaking camera.

Three friends in a pickup truck, a storm over the plains, and colossal rings covered in eyes pulling cattle into the sky. Max Bard made this 57-second found-footage horror film in Pray Studio from two 30-second Seedance 2.5 takes, all from the point of view of the one person holding the camera. Both prompts are below in full, with the mistakes that broke the first version and how each one was fixed.

Make your own video See the two prompts
Full of Eyes · two Seedance 2.5 takes · 0:57
Length
0:57
Generations in the cut
2 × 30 s
Engine
Seedance 2.5
Format
9:16

The short answer

How do you make a found-footage POV AI video?

Write the whole take as one camera’s point of view. Say “the frame IS the lens,” never name the phone or camera, and never use shot sizes like “wide shot” or “close-up.” The person filming gets a voice but no body, and the lens turns to whoever speaks. Plan the action second by second, start it in the first frame, and generate it as one or two 30-second takes in Director Mode so nothing cuts away.

The first mistake

Never write “iPhone”
in a POV prompt.

The first version of this film said the footage was shot on Cole’s phone, held in his hand. The model did exactly that: it drew a hand holding a phone in the middle of nearly every frame. Video models draw the objects you name. Describe the view, not the device.

AI video frame with a hand holding a smartphone in front of floating cattle
Before: “shot on his phone, held in his hand”
First-person AI video frame: cattle pulled into the sky outside a truck window, no phone in frame
After: “the frame IS the lens, no device is ever visible”

The same goes for the word iPhone. If you want footage that feels like it came off a phone, describe how it looks: handheld, shaking, natural light, exposure pumping on a bright sky, rain on the lens. Leave the phone out.

What fixed it

Nine lines that keep
one camera in one hand.

Every line here is copied from the prompts below. Each one is there because a draft broke without it.

In the promptWhat it does
“The frame IS the camcorder lens. We see the world directly through it, full frame.”Describes the view, not the device. The model films through the lens instead of filming a camera, so nothing shows up in front of the picture.
“No camera, phone, screen or device is ever visible in frame.”The safety net. If any word in the prompt names a device, this line tells the model it still must not draw one.
“No cuts, no edits, no other angles, no third-person view”Video models love coverage. Without this line a POV take drifts into a shot of the truck from outside, and the found-footage feeling is gone.
“When @Cole talks, his voice comes from behind the lens.”The person filming is heard and never seen. Naming him with his @tag keeps his own voice attached, so his lines come out in his voice and not someone else’s.
“When @Dez or @Maya talk, the lens is pointed at their face so we see them say it.”Every other line is spoken on camera. The viewer always knows who is talking, and the camera has a reason to swing.
“FRAME ONE IS ALREADY MID-ACTION.”On Reels and TikTok the first second decides the scroll. The cattle are already in the air on frame one, not standing in a field for three seconds first.
“It ALWAYS stays FAR AWAY, miles from the truck, high in the sky”Scale is the horror. When the creature came close in the first version it looked like a prop. Far away, it looks like weather.
“All three are terrified the whole time”Without it, AI actors drift toward a smile between lines, which kills a horror take. Spell out the fear and ban the smile.
“NO BACKGROUND MUSIC, no music, no instruments, no score.”Found footage has no soundtrack. The fear comes from wind, rain, breathing and the creature’s hum, so the music has to be ruled out in so many words.

Two more: shot-size words like “wide shot” and “close-up” make the model cut to another angle, so describe the lens moving instead (“the lens whips left to Dez”). And ask for the VHS look if you like it, but Seedance 2.5 adds only a little of it, so add grain or tape damage in your edit if you need more.

The cast

Three character sheets,
one face each, both takes.

Each person in the film is an actor made in the Actors tab from a written description, with a character sheet and a voice. Tagging @Cole, @Dez and @Maya in the prompt is what keeps the same faces and voices across both 30-second takes.

AI character sheet for Cole: a lean storm chaser in a backward red trucker cap and olive canvas jacket
@Cole Behind the camera. Never on screen, only his voice.
AI character sheet for Dez: a broad-shouldered man with a boxed beard in a navy hoodie and black rain shell
@Dez The driver, the one holding it together.
AI character sheet for Maya: a young woman with a loose dark braid in a mustard-yellow rain jacket
@Maya In the back seat. She sees it turn toward them first.

Copy them

The two found-footage prompts,
in full.

Each prompt is one 30-second Seedance 2.5 take in Director Mode, 9:16. @Dez, @Maya and @Cole are actors with their own faces and voices. @Image4 to @Image7 are four reference pictures of the Ophanim.

Take 1: The truck and the barn

Cattle already in the air on frame one, the drive, Maya spotting the barn, and the door sliding shut.

The full prompt · Seedance 2.5 · 30 seconds · 9:16

NO BACKGROUND MUSIC, no music, no instruments, no score. Audio is natural sound effects and dialogue only. Vertical 9:16. Dark and dreary skies, dark green-black storm clouds like a coming tornado, late afternoon light that is almost dusk, wind and rain.

LOOK, EVERY FRAME: this is a degraded 1996 home video recorded on a cheap VHS camcorder and played back from a worn tape. Every single frame shows heavy VHS damage: soft blurry low-resolution picture, visible horizontal scan lines, red and blue color bleeding on every edge, smeared washed-out colors, grey crushed blacks, glowing blown-out highlights, a thin rolling tracking-noise band along the bottom edge, tape wobble, brief white dropout specks. It must never look like clean modern digital video or a movie. Violently shaky, panicked handheld movement. No on-screen text.

POINT OF VIEW, THE WHOLE TAKE: one unbroken 30 second first-person recording. The frame IS the camcorder lens. We see the world directly through it, full frame. No camera, phone, screen or device is ever visible in frame. No cuts, no edits, no other angles, no third-person view, never a view from outside the truck. The person filming is @Cole, in the front passenger seat. @Cole is never seen in this take, not his face, not his head, not his cap, not his shoulder. When @Cole talks, his voice comes from behind the lens. When @Dez or @Maya talk, the lens is pointed at their face so we see them say it.

PEOPLE: exactly three people, all inside one old pickup truck. @Dez drives in the front left seat. @Cole films from the front passenger seat. @Maya stays in the BACK seat behind them the whole drive. All three are terrified the whole time: wide eyes, heavy breathing, shaking, crying. Nobody ever smiles or laughs.

THE CREATURE: body, color and texture exactly like @Image4 and @Image5: dark, dead, decrepit, charred grey bone and black rusted iron, wet rotting organic flesh, pitted and cracked like an ancient shipwreck. It is NOT gold, NOT shiny, NOT polished, NOT clean metal. Its shape and scale like @Image6 and @Image7: a colossal Ophanim made of THREE separate gigantic circular rings passing through each other at different angles like a gyroscope, each full circle visible and each one slowly turning on its own axis. Dozens of huge wet human eyes are set along every rim, blinking out of sync. It ALWAYS stays FAR AWAY, miles from the truck, high in the sky, half inside the storm clouds. It never comes down to the ground, never comes near the truck or the people, and never fills more than the top third of the frame. Directly beneath it a huge tornado-like vortex of dust, hay, fence boards and cattle spirals up from the fields into its center. Its sound rolls across the plains: a very deep, slow sub-bass hum, the heavy metallic clanking and grinding of the turning rings, and wet organic creaking, under the roar of the wind.

Scene Action, second by second:
0-1s: FRAME ONE IS ALREADY MID-ACTION. Nothing is calm, nothing stands still. The lens looks out the open passenger window of the speeding truck, the side mirror at the bottom edge. A dozen black and brown cattle are ALREADY high in the air, five to ten meters off the ground, legs kicking, spinning, bellowing, being sucked upward. Hay and dust spiral up with them. Roaring wind. No cow is standing on the ground.
1-3s: The lens tilts up following the flying cattle. They are pulled up into a huge swirling column of dust and debris that climbs miles into the clouds.
3-6s: At the top of the column, far away and high in the clouds, the dark decrepit creature: three gigantic interlocking circular rings slowly turning through each other, eyes along every rim, the deep hum and metallic clanking rolling across the fields. @Cole, behind the lens, panicking: "Dez. DEZ. Drive! Drive drive drive!"
6-9s: The lens whips left to @Dez at the wheel, his face filling the left half of the frame: pale, sweating, eyes wide, glancing up at the sky then back at the road, both hands gripping the wheel. @Dez shouts, terrified: "Where?! Where do I go?!" Gravel roars under the tires.
9-12s: The lens swings forward to look out the windshield: the gravel road racing toward us, wipers slapping rain, power lines whipping, a fence post rips out of the ground and flies upward across the road ahead.
12-16s: The lens twists around between the front seats to the back seat. @Maya sits in the back seat pressed against the rear window, looking back and up at the sky behind the truck, crying, one hand over her mouth. She turns to the lens and says, shaking: "It's turning... it's turning toward us." Behind her, through the rear window, far away in the sky, the three dark rings slowly turn and all the eyes along their rims shift in the same direction.
16-19s: The lens looks past @Maya out the rear window: far away, the distant rings slowly tilt toward the road, the debris vortex bending toward us, the hum growing louder.
19-22s: @Maya leans forward from the back seat between the front seats, points ahead through the windshield and screams: "The barn! Get in the barn!" The lens swings forward: an old weathered wooden barn beside the road a hundred meters ahead.
22-25s: The truck skids off the road onto the grass beside the barn. The passenger door flies open. The lens jolts out into the rain and runs: wet grass and mud bouncing below, @Dez and @Maya sprinting upright on their feet ahead of us toward the open barn doorway.
25-28s: Inside the barn, @Dez and @Maya drag the big sliding door shut. Darkness, thin slits of grey light between the wall boards, everyone breathing hard, rain hammering the roof.
28-30s: The lens presses to a gap between two wall boards. Through it, across the field, far away in the sky, the three dark rings turn slowly in the clouds. @Cole, behind the lens, whispering: "Nobody move."

Ending on the far distant rings seen through a gap in the barn wall. The creature stays far away in the sky for the entire take. @Cole is never visible.
Make this take in Create

Opens the Create tab with this prompt, 30 seconds and 9:16 set, and Dez, Maya and the four Ophanim pictures attached as references.

Take 2: The roof, the run and the culvert

The barn roof tears away, they run for a drainage culvert, and the tape tears into static.

The full prompt · Seedance 2.5 · 30 seconds · 9:16

NO BACKGROUND MUSIC, no music, no instruments, no score. Audio is natural sound effects and dialogue only. Vertical 9:16. Dark and dreary skies, dark green-black storm clouds like a tornado, dusk, heavy wind and pouring rain.

LOOK, EVERY FRAME: this is a degraded 1996 home video recorded on a cheap VHS camcorder and played back from a worn tape. Every single frame shows heavy VHS damage: soft blurry low-resolution picture, visible horizontal scan lines, red and blue color bleeding on every edge, smeared washed-out colors, grey crushed blacks, glowing blown-out highlights, a thin rolling tracking-noise band along the bottom edge, tape wobble, brief white dropout specks. It must never look like clean modern digital video or a movie. Violently shaky, panicked handheld movement. No on-screen text.

POINT OF VIEW, THE WHOLE TAKE: one unbroken 30 second first-person recording, continuing a moment later. The frame IS the camcorder lens, held at eye level by @Cole. We see the world directly through @Cole's eyes, full frame. No camera, phone, screen or device is ever visible in frame. No cuts, no edits, no other angles, no third-person view. @Cole is NEVER visible anywhere in this take: not his face, not the back of his head, not his red cap, not his shoulder, not his arm. Only his voice is heard. Only TWO people are ever in frame: @Dez and @Maya. When @Cole talks, his voice comes from behind the lens. When @Dez or @Maya talk, the lens is pointed at their face so we see them say it.

PEOPLE: @Dez and @Maya, the same clothes as before, soaked from the rain, in front of the lens. All of them are terrified the whole time: wide eyes, heavy breathing, shaking, crying. Nobody ever smiles or laughs.

THE CREATURE: body, color and texture exactly like @Image4 and @Image5; shape and scale like @Image6 and @Image7. A colossal Ophanim made of THREE separate gigantic circular rings passing through each other at different angles like a gyroscope, each one slowly turning on its own axis, decrepit dark cracked bone fused with corroded rust-black iron and wet organic flesh, dozens of huge wet human eyes along every rim. It ALWAYS stays FAR AWAY, miles off, high in the sky, half inside the storm clouds. It never comes down to the ground, never comes near the people, and never fills more than the top third of the frame. We never see a single eye up close. Beneath it a huge tornado-like vortex of dust, hay, boards and debris spirals up from the fields into its center. Its sound: a very deep, slow sub-bass hum, heavy metallic clanking and grinding of the turning rings, wet organic creaking, under the roar of wind and rain.

Scene Action, second by second:
0-3s: Inside the dark barn, the lens is crouched low behind a wooden stall looking at @Dez and @Maya, who crouch in the straw in front of us, soaked, shaking. The deep hum swells. Loose straw and dust lift off the floor and spiral up toward the rafters. Nails rattle in the boards. Old tools swing on their hooks.
3-6s: The lens moves in on @Dez's face, close: sweat, rain, wide eyes looking up at the roof. @Dez whispers to the lens, terrified: "It's getting closer."
6-10s: The lens presses to a gap in the wall boards. Through the gap, across the dark field, far away in the sky: the three rings, still miles away, turning slowly in the clouds above the debris vortex. Far off on the horizon a small white farmhouse is being pulled apart, its roof and boards spiraling up into the vortex.
10-13s: Above them the barn roof starts tearing open: boards rip loose one after another and fly up into the storm, rain and wind pour in. @Maya screams. The lens tilts up at the opening roof: a strip of churning grey sky with debris flying through it.
13-15s: @Cole shouts from behind the lens, only his voice: "Back door. Go. GO!" The lens swings to a small back door. @Dez kicks it open into the rain.
15-19s: Running out the back into pouring rain across a muddy field. The lens shakes violently. @Dez and @Maya run just ahead of us, wind tearing at their clothes, hay and leaves whipping past and flying upward.
19-22s: They slide down a muddy embankment and scramble into a large round concrete drainage culvert under the road. The lens follows them into the dark concrete pipe, water running around their feet.
22-25s: Inside the culvert, @Maya and @Dez press against the curved concrete wall facing the lens, soaked and shaking. @Maya grabs @Dez's arm, stares into the lens and whispers, crying: "Don't look at it. Don't look at it."
25-28s: The lens turns to the round opening of the culvert. Through it, across the field, the barn they just left lifts off the ground in pieces and spirals up into the sky. Far beyond it, high in the clouds and still miles away, the three colossal rings turn, and every eye on every ring rolls at once to stare toward the culvert. The hum deepens into a roar.
28-30s: The VHS tape tears: the picture rolls and jitters, tracking bars sweep across it, the colors smear, then the image breaks into grey snow static.

Ending on VHS static. The creature stays far away in the sky for the entire take. Only @Dez and @Maya are ever seen.
Make this take in Create

Opens the Create tab with this prompt, 30 seconds and 9:16 set, and Dez, Maya and the four Ophanim pictures attached as references.

Make it yours: keep the look, point of view and people blocks, change the creature and the second-by-second action. Paste it into Director Mode ↗

The edit

57 seconds, cut from
the best of five drafts.

Two shots, five 480p drafts between them. The cut takes each moment from whichever draft got it right, so it still plays as one recording.

  1. 0:00

    Cattle already in the air outside the truck window. Cole: “Dez. Drive!”

  2. 0:06

    The lens whips to Dez at the wheel: “Where do I go?!”

  3. 0:12

    Back seat. Maya, crying: “It’s turning toward us.” The rings turn in the rear window.

  4. 0:17

    A jump cut straight to Maya’s “The barn! Get in the barn!” It plays like the recording stopping and starting again, and it drops two seconds that looked like the truck reversing toward the creature.

  5. 0:20

    Out of the truck and running for the barn, from a second take of the same shot.

  6. 0:25

    The barn door slides shut. Cole, whispering through a gap in the boards: “Nobody move.”

  7. 0:28

    Take 2 begins: Dez and Maya crouched in the straw as it lifts off the floor.

  8. 0:30

    A whip pan onto Dez’s face, “It’s getting closer,” taken from another draft. The keeper draft jumped from wide to close here, which no handheld camera can do.

  9. 0:34

    Through the wall boards, the far rings pull a farmhouse apart. Then the roof goes.

  10. 0:40

    Out the back door, across the mud and into a concrete culvert.

  11. 0:49

    Maya, to the lens: “Don’t look at it. Don’t look at it.”

  12. 0:52

    The barn lifts away into the sky. Every eye turns toward them. The tape tears to static.

Make your own

How to make a found-footage film
in Director Mode.

Swap the Ophanim for your own story: a storm, a flood or a stranger at the door. The method stays the same.

  1. 01

    Cast your people, with voices

    Make each character an actor in the Actors tab with a character sheet and a voice. Tag them by name in the prompt (@Dez, @Maya, @Cole) and they keep the same face and voice in both takes.

  2. 02

    Upload pictures of what you are filming

    Attach two to four reference pictures of what you are filming to every shot. With no references the first version of this film drew a cartoon. Say which pictures set the color and texture and which set only the shape.

  3. 03

    Start a Director Mode film at 9:16 with two 30-second shots

    One 30-second take per sequence keeps the camera, the people and the light continuous. Four short takes cut together never feel like one recording.

  4. 04

    Write each take as one camera, second by second

    Look, point of view, people, creature, then the action in two or three second beats with every line of dialogue in quotes. Both full prompts are on this page.

  5. 05

    Make 480p drafts and watch every one

    Drafts are cheap. Check each one for a device in frame, the camera holder showing up, a smile, a cut or a zoom, and re-roll the take that breaks. This film used five drafts for its two takes.

  6. 06

    Cut the best seconds of every draft, then finalize at 1080p

    In Post-Production, take each moment from whichever draft got it right. Finalize only the drafts the cut uses.

Questions

Questions about found-footage AI video.

How do you make a found-footage AI video?
Write the prompt as one unbroken first-person recording: the frame is the lens, no device is ever in frame, no cuts and no other angles. Give the person filming a voice but no body, plan the action second by second, and generate one 30-second take per sequence on Seedance 2.5 in Director Mode in Pray Studio.
Why does my AI POV video show a hand holding a phone?
Because the prompt names the phone. Video models draw every object you mention, so “filmed on his iPhone” or “phone held in his hand” puts a phone in the middle of the shot. Describe the view instead: “the frame IS the lens, no camera, phone, screen or device is ever visible in frame.”
Can AI make a video that looks like it was shot on an iPhone?
Yes, but describe how the footage looks, not the phone: handheld, shaky, natural light, exposure pumping on bright sky, autofocus hunting, rain on the lens. Leave the word iPhone out of the prompt, or the model may draw one.
Why does my first-person AI video cut to other angles?
Shot-size words like “wide shot” and “close-up” tell the model to cut to coverage. Describe movement as the view moving (“the lens whips left to Dez”) and add “no cuts, no edits, no other angles, no third-person view.”
How long can one found-footage AI take be?
Seedance 2.5 in Director Mode makes takes up to 30 seconds. This 57-second film is two 30-second takes, trimmed and cut together.
How do you get the person filming to talk without showing up?
Tag him once in each prompt and say his voice comes from behind the lens and he is never seen. In the second take of this film he still showed up in frame until the prompt listed his face, his cap, his shoulder and his arm, and said only two people are ever in frame.
Can I make a found-footage AI video in 1:1 square?
Director Mode makes 16:9, 9:16 and 21:9. This film is 9:16 for Instagram Reels and TikTok.
What is an Ophanim?
In Ezekiel 1 the prophet sees wheels beside the living creatures, like a wheel inside a wheel, with rims full of eyes. Ophanim is the Hebrew word for those wheels. This film is an artistic interpretation.
Four Ophanim images, with every prompt ↗How to write AI video prompts ↗See current plans and credits ↗

Make your own

Your story.
One camera. Two takes.

Cast your people, paste a prompt from this page, and make your first 480p draft in Director Mode.

Make a found-footage film

Pray Studio is made by Pray.com.