The Gangnam Style AI trend, with you and your two best friends

How to make a Gangnam Style AI video yourself

You do not need this page to make one. This is how the video is put together, with the prompts for every shot, so you can build it with any picture model and image-to-video model you like.

1. Step 1

   Draw one still per shot from the three photos

   Use a picture model that takes reference photos. The two selfies need all three photos; each solo needs only that friend's photo, and the close-up only the first. Describe the street the same way in every still, the office towers, the white lamps and the traffic, so the place does not change between shots.

2. Step 2

   Say who is where, left to right

   Name each person by the order you sent the photos and say the order across the frame: image 2 on the left, image 1 in the middle, image 3 on the right. Say too that every face and every outfit comes from its own photo, or the model trades them between friends.

3. Step 3

   Animate each still, then cut on the bar

   Give each still to an image-to-video model with the moves spelt out: wrists crossed as if holding reins, bouncing foot to foot in a skipping, galloping step. Pick a beat of 120 a minute, trim each clip to two bars (four seconds) and the last to one, lay them end to end at 9:16 with hard cuts, and add the song on TikTok.

The shots, second by second

The prompts, shot by shot

Each shot is made in two steps. Give its still prompt to a picture model with the photos listed for that shot as references, then give the picture it draws to an image-to-video model with the motion prompt. Add these endings to every prompt: they set the look and keep the faces the same from shot to shot.

1. 0–4sThe selfie

   Photos to attach, in orderImage 1: The one filming, in the middleImage 2: On the leftImage 3: On the right

   Still prompt

   Copy

   A tight selfie shot of three friends' heads and shoulders squeezed together cheek to cheek, filling the whole frame, taken at arm's length from just above eye level by the one in the middle. Their arm is stretched out of the frame toward the lens, and nothing they hold is in the picture. From left to right: the person in image 2, the person in image 1, the person in image 3. All three have their eyes open. The person in image 1, in their own clothes from the photo, is in the middle, nearest the lens, eyes open, smiling with lips closed, face in full view. The person in image 2, in their own clothes from the photo, leans in on the left of the frame, eyes open, smiling, face in full view. The person in image 3, in their own clothes from the photo, leans in on the right of the frame, eyes open, smiling, face in full view. Behind them, tall glass office towers with lit windows and white street lamps at night are softly out of focus into round city lights, with no sign, logo, print or writing anywhere. Each face and each outfit comes from that person's own photo alone, never a mix of the three: the one in the middle has exactly the face, gender, age and clothes of the person in image 1. Exactly these three people are in the picture; nobody else is near them, and the street behind them is empty of people.

   Motion prompt

   Copy

   The three friends bounce together to a pumping dance beat, laughing and grinning into the lens, bobbing their heads and shoulders on every beat, and the two at the sides throw an arm up in the air and cheer. Everyone stays in their place, the one in the middle filming at arm's length. The camera shakes a little with them, held at arm's length. They do not sing or mouth any words.

2. 4–8sSolo one

   Photo to attachOn the left

   Still prompt

   Copy

   The person in the reference photo, in their own clothes from the photo, stands on the pavement facing the camera, wrists crossed in front of their chest, knees a little bent, ready to dance, grinning, face in full view. A full-body shot from about three metres away, seen head to toe, shoes on the pavement, as a friend films them. Behind them is a wide city pavement at night in a modern business district: tall glass office towers with rows of lit windows rising behind, tall white street lamps, a few bare trees, and the red and white streaks of passing traffic on the road beside them, under a dark sky. None of it has any sign, billboard, flag, logo, print or writing on it: the buildings, the shop fronts and the lamp posts are plain. Only this one person is in the picture; nobody else is near them, and the street behind them is empty of people.

   Motion prompt

   Copy

   They dance on the spot to a pumping dance beat: wrists crossed in front of their chest as if holding a pair of reins, they bounce from foot to foot in a skipping, galloping step, knees lifting high, shoulders bouncing with every beat, grinning at the lens. The camera holds them head to toe, handheld. They do not sing or mouth any words.

3. 8–12sSolo two

   Photo to attachOn the right

   Still prompt

   Copy

   The person in the reference photo, in their own clothes from the photo, stands on the pavement facing the camera, one hand raised above their head, the other held out in front of their chest, knees a little bent, ready to dance, grinning, face in full view. A full-body shot from about three metres away, seen head to toe, shoes on the pavement, as a friend films them. Behind them is a wide city pavement at night in a modern business district: tall glass office towers with rows of lit windows rising behind, tall white street lamps, a few bare trees, and the red and white streaks of passing traffic on the road beside them, under a dark sky. None of it has any sign, billboard, flag, logo, print or writing on it: the buildings, the shop fronts and the lamp posts are plain. Only this one person is in the picture; nobody else is near them, and the street behind them is empty of people.

   Motion prompt

   Copy

   They dance on the spot to a pumping dance beat, bouncing from foot to foot in a skipping, galloping step, knees lifting high, the raised hand circling above their head as if swinging a lasso and the other held out in front, grinning at the lens. The camera holds them head to toe, handheld. They do not sing or mouth any words.

4. 12–16sBack to you

   Photo to attachThe one filming, in the middle

   Still prompt

   Copy

   A close, slightly wide-angle selfie shot taken at arm's length from just above eye level, the arm holding the camera reaching out of the bottom corner of the frame, the hand itself out of frame. The person in the reference photo, in their own clothes from the photo, is alone in the frame, their head and shoulders filling most of it, eyes open, smiling with lips closed, face in full view. Behind them, tall glass office towers with lit windows and white street lamps at night are softly out of focus into round city lights, with no sign, logo, print or writing anywhere. Only this one person is in the picture; nobody else is near them, and the street behind them is empty of people.

   Motion prompt

   Copy

   They grin into the lens, bob their head and shoulders to the pumping beat and wave at the camera with their free hand, laughing. The camera shakes a little with them, held at arm's length. They do not sing or mouth any words.

5. 16–18sAll three again

   Photos to attach, in orderImage 1: The one filming, in the middleImage 2: On the leftImage 3: On the right

   Still prompt

   Copy

   A tight selfie shot of three friends' heads and shoulders squeezed together cheek to cheek, filling the whole frame, taken at arm's length from just above eye level by the one in the middle. Their arm is stretched out of the frame toward the lens, and nothing they hold is in the picture. From left to right: the person in image 2, the person in image 1, the person in image 3. All three have their eyes open. The person in image 1, in their own clothes from the photo, is in the middle, nearest the lens, eyes open, beaming with lips closed, face in full view. The person in image 2, in their own clothes from the photo, leans in on the left of the frame, beaming, eyes open, face in full view. The person in image 3, in their own clothes from the photo, leans in on the right of the frame, beaming, eyes open, face in full view. Behind them, tall glass office towers with lit windows and white street lamps at night are softly out of focus into round city lights, with no sign, logo, print or writing anywhere. Each face and each outfit comes from that person's own photo alone, never a mix of the three: the one in the middle has exactly the face, gender, age and clothes of the person in image 1. Exactly these three people are in the picture; nobody else is near them, and the street behind them is empty of people.

   Motion prompt

   Copy

   The three friends bounce together on the last beat, the two at the sides crossing their wrists in front of their chests and throwing their heads back laughing, then all three grin into the lens. Everyone stays in their place, the one in the middle filming at arm's length. The camera shakes a little with them, held at arm's length. They do not sing or mouth any words.

Where it goes wrong

• Keep the selfie tight

  Ask for an arm's-length selfie and a picture model often draws the arm itself, a huge hand in the foreground with the three of you small behind it, and with small faces it loses track of whose is whose. Ask for heads and shoulders filling the frame, the arm out of the frame, and the street behind as a blur.

• Do not name what is held

  A picture model draws what you name. Say the person in the middle is holding a camera and you may get a big black camera in their hand; say phone and you get a phone. Say instead that their arm reaches out of the frame toward the lens and nothing they hold is in the picture.

• Describe the dance, not the horse

  Everyone calls it the horse dance, but a picture model told "horse" or "riding" may draw a horse. Describe what the body does: wrists crossed in front, a skipping, galloping step, one hand circling overhead like a lasso.

• Leave the signs off the street

  A city street at night is full of shop signs, billboards and flags, and a picture model fills them with lettering and logos. Ask for plain buildings and lamp posts with no sign, flag, logo or writing anywhere.

• Only people who said yes

  Use your own photos or those of friends who agreed. The first versions starred real athletes; do not use celebrities, public figures, the people in the original clip, or anyone who has not agreed to be in the video: a realistic video of a real person made without their consent can hurt them, and platforms take it down.

• Label it as AI-generated

  Switch on TikTok's AI-generated content label before you post. A realistic clip of friends dancing in the street is exactly the kind people ask about, and viewers trust a post that says so.