First-Person POV AI Video Prompts: 10 Shots That Feel Real

You type “POV: you are a barista on your first day” into an AI video model, and it gives you a nice third-person shot of a barista. Smiling. Facing the camera. Exactly the opposite of what you asked for.

First-person POV is one of the most useful formats in short-form video right now. The “POV: you are…” caption trend, GoPro-style action clips and hands-in-frame cooking or crafting shots all put the viewer inside the scene instead of watching it. The trouble is that AI video models are trained on a lot more footage of people than footage from people, so they drift back to third person unless you box them in.

This guide gives you the prompt structure that keeps the camera behind the eyes, 10 copy-ready first-person POV prompts, the hand rules that stop extra fingers from showing up, and notes for Veo 3.1, Kling 3.0, Seedance 2.0 and the rest.

Why AI Models Keep Breaking First-Person POV

Three things go wrong again and again:

The model shows you the person. “POV of a firefighter” contains the word firefighter, and the model wants to draw one. Without an explicit rule that the camera wearer is never visible, you get a portrait.
The hands get weird. A first-person shot almost always has hands in the lower third of the frame. Hands are still a weak area for video models, and two hands doing two different things doubles the chance of a sixth finger or a hand that melts into a cup.
The motion is wrong. Real POV footage has a specific motion signature: a head-mounted camera turns with the head, a chest-mounted camera bobs with each step and stays level with the torso, a phone held in front of you sways with the arm. A generic “moving camera” looks like a drone, not a person.

Every prompt below solves these three problems on purpose.

The First-Person POV Prompt Formula

Build every POV prompt from five parts, in this order:

Perspective lock. Open with “First-person POV shot” and state the mount: head-mounted action camera, chest-mounted camera, helmet cam, or eye-level POV. Google’s own Veo documentation uses this phrasing in an example (“A POV shot from a vintage car driving in the rain, Canada at night, cinematic”), and every major model understands it.
What is visible of “you.” Name exactly what the viewer sees of their own body: “only the viewer’s hands and forearms visible at the bottom of the frame,” or “no part of the camera wearer is visible.”
The action, from the wearer’s side. Write the action as what the hands do or where the head turns: “the right hand pulls a lever,” “the view turns left toward the door.” Avoid “the chef chops onions,” which invites a shot of a chef.
The world reacting to the camera. Other characters looking into the lens is the strongest POV signal there is. “A customer looks directly into the camera and says…” instantly reads as first person.
Lens, motion and sound. Wide-angle or slight fisheye for action cams, natural head sway for walking, and sound from the wearer’s position: breathing, footsteps, wind noise on the mic.

This is the same six-part prompt structure the rest of the site uses, with the camera position promoted to the very first words. If you want more vocabulary for the motion itself, the camera movement cheatsheet covers pans, tilts and tracking moves you can combine with a POV mount.

Mount Vocabulary That Changes the Output

Mount	What to write	What it looks like
Head-mounted	”head-mounted action camera, view turns with the head”	Fast turns, horizon tilts slightly with the head
Chest-mounted	”chest-mounted camera, view level with the torso, hands reach into frame”	Rhythmic bob when walking, arms enter from the bottom
Helmet cam	”helmet-mounted camera, visor edge faintly visible at the top of frame”	Motorbike, bike, ski footage
Eye-level POV	”eye-level first-person view, natural head sway”	The cinematic “you are here” look, less fisheye
Held phone	”phone held at chest height, first-person view, slight hand shake”	Casual vlog POV, pairs with a handheld phone video style
Dashboard	”dashboard POV through the windshield”	Driving clips, close to dashcam footage

Pick one mount per prompt. “Head-mounted chest camera” confuses the model about where the lens is.

10 Copy-Ready First-Person POV Prompts

Swap the details, keep the structure. Each prompt is built for a 5 to 10 second clip.

1. POV: your first shift as a barista

First-person POV shot, eye-level, natural head sway. Only the viewer's hands and forearms are visible at the bottom of the frame, wearing a green apron strap across one wrist. The hands tamp espresso grounds into a portafilter while the view glances up at a long line of impatient customers. A woman at the front of the line looks directly into the camera and says, "Is it always this slow?" Warm cafe lighting, slight handheld shake, sound of the grinder, milk steamer hiss and cafe chatter.

2. Mountain bike descent (chest cam)

First-person POV from a chest-mounted action camera on a mountain biker, wide-angle lens with slight fisheye distortion. Gloved hands grip the handlebars at the bottom of the frame. The bike rolls fast down a narrow dirt trail through pine trees, rhythmic bumps shake the view, roots and rocks rush toward the lens. Dappled afternoon sunlight through the trees. Sound of tires on gravel, heavy breathing and wind noise on the mic.

3. POV: you are a cat on the kitchen counter

First-person POV at cat height, camera about 30 centimeters above a marble kitchen counter, wide-angle lens. No part of the cat is visible except a single orange paw that reaches slowly into frame from the bottom and taps a glass of water toward the edge. A man in the background turns around, sees the camera, points directly at the lens and shouts, "Don't you dare." Bright morning kitchen light, the glass wobbles, sound of the paw tapping and a sharp gasp.

Pet POV clips like this one pair well with the ideas on the cat video prompts page.

4. Hands-in-frame cooking, top-down

First-person POV looking down at a wooden cutting board, eye-level with slight downward tilt. Only the viewer's two hands are visible: the left hand holds a red onion steady, the right hand slices it with a chef's knife in slow, even strokes. Onion layers fall apart cleanly. Soft window light from the left, shallow depth of field on the knife edge. Crisp knife sounds on the board, a pan sizzling off-screen.

5. Night delivery rider in the rain (helmet cam)

First-person POV from a helmet-mounted camera on a scooter delivery rider at night, the edge of the helmet visor faintly visible at the top of the frame. Hands in black rain gloves hold the handlebars, a delivery app glows on a phone mounted between them. The scooter weaves through a wet city street, neon signs smear across puddles, rain streaks across the lens. Sound of rain on the helmet, a muffled engine and a car horn in the distance.

6. POV: you open the door and everyone yells “surprise”

First-person POV, eye-level, handheld phone feel. The viewer's right hand pushes open an apartment door into a dark living room. The lights snap on and six friends jump up from behind a sofa, all looking directly into the camera, and shout "Surprise!" at the same time. Confetti falls across the lens. Warm tungsten lighting, slight camera jolt backward in shock. Sound of the door creak, then a loud cheer and a party popper.

7. Rooftop edge (vertigo shot)

First-person POV, head-mounted action camera, view turns with the head. The camera wearer stands on the edge of a skyscraper rooftop, sneaker toes visible at the bottom of the frame. The view slowly tilts down past the shoes to the busy street far below, then snaps back up to the horizon. Golden hour light, strong wind. Sound of wind gusts across the mic and faint city traffic far below.

8. POV: new teacher, first class

First-person POV, eye-level, natural head sway. The viewer walks into a classroom holding a coffee mug in the right hand, visible at the bottom right of the frame. Twenty-five teenagers at their desks turn to stare directly into the camera at the same moment, the room goes silent. One student in the back row slowly raises a hand. Flat fluorescent classroom lighting, slight handheld shake. Sound of footsteps on linoleum, then total silence, then a single chair squeak.

9. Pottery wheel, hands only

First-person POV looking down at a spinning pottery wheel, chest-mounted camera. The viewer's wet, clay-covered hands press into a lump of grey clay and slowly pull it upward into a tall cylinder. Water glistens on the clay surface. Soft studio skylight from above, shallow depth of field. Sound of the wheel motor humming and wet clay squelching.

Hands-in-frame craft shots like this also do well as oddly satisfying videos.

10. Horror hallway, flashlight

First-person POV, eye-level, shaky handheld feel. The viewer's right hand holds a small flashlight at the bottom of the frame, its beam the only light in a long, dark hospital corridor. The beam sweeps left across peeling green walls, then right, where a wheelchair sits at the far end. The wheelchair slowly rolls one meter toward the camera on its own. Cold blue tint, heavy film grain. Sound of shallow breathing, a dripping pipe and a faint wheel squeak.

Hands in Frame: The Rules That Stop Glitches

Most failed POV clips fail at the hands. These rules keep them usable:

Say which hand does what. “The right hand slices, the left hand holds the onion steady” gives the model a clear job for each hand. “Hands chopping” leaves it guessing.
One hand busy, one hand still. Two hands doing complex, different actions at once is where fingers merge. Keep one hand holding or resting.
Give the hands clothing. Gloves, a sleeve cuff, a watch or an apron strap anchor the hands to a body and make them read as “yours.”
Keep the object simple. A mug, a knife, a flashlight, handlebars. Tiny objects with small parts (keys, jewelry clasps, phone keyboards) are harder.
Anchor them to the bottom of the frame. “Visible at the bottom of the frame” keeps hands where a real POV camera sees them, instead of floating mid-frame.

If a clip keeps producing warped hands, try a negative prompt on models that support one (Kling has a dedicated field). The negative prompts guide covers what to put in it.

Model Notes for POV Clips

Veo 3.1. Veo handles the “world reacts to the lens” trick well because it generates dialogue and sound natively. Google’s docs recommend putting speech in quotes, so write “A woman looks into the camera and says, ’…’” and she will address the viewer. Veo clips run 4, 6 or 8 seconds, so plan one POV beat per clip. More phrasing ideas are in the Veo 3 prompts guide and on the Veo 3 prompt generator page.

Kling 3.0. Strong on physical motion, which matters for bike and helmet clips where the camera shakes with the terrain. Kling 3.0 generates clips from 3 to 15 seconds and has a multi-shot mode; for POV, keep every shot in first person or the model may cut to a reaction shot of the wearer. See the Kling AI prompt guide and the Kling AI prompt generator.

Seedance 2.0. Seedance accepts reference videos as well as images. If you have a real GoPro clip with the motion you want, upload it and tell the prompt to copy the camera motion from that video while your text describes the new scene. The Seedance 2.0 prompt guide explains the @Video1 style tags, and the Seedance prompt generator page has more starting points.

Hailuo. MiniMax’s models respond well to short, direct motion language. Lead with “First-person POV” and keep the prompt to one action. The Hailuo AI prompt generator and the MiniMax Hailuo prompt guide go further.

Runway Gen-4.5. Runway’s own prompting guide advises starting simple and adding one element at a time. For POV, that means: perspective lock plus one action first, then add dialogue and sound once the perspective holds.

Fixing the Three Most Common POV Failures

The camera flips to third person halfway through. Repeat the lock at the end of the prompt: “The camera wearer is never visible; the entire clip stays in first person.” Also check for job titles in the action sentence. “The surgeon’s hands” is safer than “the surgeon.”

The fisheye is too strong. “Wide-angle lens with slight fisheye” is enough for action cams. Plain “fisheye lens” often gives you a circular, warped image. For cinematic POV, drop the lens note and write “eye-level first-person view.”

The motion feels like a drone. Add a body cue: “natural head sway,” “rhythmic walking bob,” “slight hand shake.” Real people do not glide.

Turning “POV: You Are…” Ideas Into a Batch

The fastest way to test a POV concept is to try several versions of it at once, because the first idea is rarely the one that performs. This is what I built LzyPrompt for. Type the idea in plain words, like “POV: you are a new lifeguard and a kid keeps waving at you,” set the Camera Style dropdown if you want a specific look (handheld phone video, dashcam POV or selfie shot are good matches for POV clips), and you get 10 prompt variations. Each one has a scene title, the full prompt with action beats and sound cues, and a ready-to-post caption. Paste them into Veo, Kling, Seedance or whatever model you use, or export all 10 to CSV if you generate in bulk.

For more short-form formats built around the viewer’s eye line, look at the selfie vlog prompts and the TikTok and Reels prompt guide.

FAQ

How do I stop the AI from showing the person in a POV video?

Open with “First-person POV shot,” say exactly which body parts are visible (“only the hands and forearms at the bottom of the frame”), and add a closing line like “the camera wearer is never visible.” Avoid naming the wearer’s job as the subject of the action sentence.

What is the best camera phrase for a GoPro look?

“Head-mounted action camera” or “chest-mounted action camera,” plus “wide-angle lens with slight fisheye distortion.” Add a body motion cue such as “rhythmic bumps shake the view” for sports clips.

Can other characters talk to the viewer in a POV clip?

Yes, and it is the strongest way to sell first person. On models with native audio, such as Veo 3.1 and Kling 3.0, write “looks directly into the camera and says” followed by the line in quotes. Keep the line short enough to fit the clip.

Why do the hands look wrong in my first-person clips?

Usually because both hands are doing different complex actions or holding small, detailed objects. Give one hand a simple job, keep the other still, add gloves or sleeves, and anchor the hands to the bottom of the frame.

Start Your First POV Batch

First-person POV works when the prompt leaves the model no room to step outside the wearer’s head: lock the perspective first, say what the viewer sees of their own body, write the action from the hands, and let the world look back into the lens.

If you want 10 variations of your next “POV: you are…” idea in one go, try LzyPrompt free for 7 days. Plans are $9 a month or $60 a year, with unlimited prompts.