How to Do the Gangnam Style AI Trend
The Gangnam Style AI trend copies one TikTok: three friends on a Gangnam street at night, singing PSY's "Gangnam Style" into a phone that whips from face to face for close-up solos, and ending together under the Korean flags. Since mid-October 2026 people have remade it with their own faces, with Harry Potter characters and with their dogs. This guide covers where it comes from, every shot, the photos you need, the one-photo way, friends and pets, the do-it-yourself method and what goes wrong.
What the Gangnam Style AI trend is
It is not the horse dance. On 5 October 2026 three friends filmed a selfie video at night on a street in Gangnam, Seoul, singing along to PSY's "Gangnam Style". The phone starts on all three, whips to each of them for a solo, a pout, a hair flip, hands in the hair, and ends on the trio laughing under the flags. @emiliyyam's clip passed 18.7 million plays in three weeks.
From 12 October people remade it with AI. A Russian tutorial by @bikonyavamm passed 3.2 million plays; @yapayzekaserisi put Harry, Ron and Hermione in the snow at Hogwarts, then Gandalf, Frodo and Aragorn below a white city. CapCut templates followed in Arabic, Russian, Vietnamese and Spanish.
The 21 seconds, shot by shot
The original is one continuous handheld take with no hard cuts. The camera moves between the friends with fast whip pans, and the order never changes:
| Time | Who is on screen | What happens |
|---|---|---|
| 0–2.5 s | All three | Squeezed into one selfie frame under the Korean flags, singing |
| 2.5–4.7 s | The friend in the middle | A close-up solo and a pout to the lens |
| 4.7–8.3 s | The friend on the right | After a whip pan: a hair flip, hands through the hair, by the road |
| 8.3–12.1 s | The friend on the left | After a whip pan: a solo in front of a lit glass lobby |
| 12.1–16 s | The friend in the middle | The strongest whip pan, then a second solo |
| 16–21 s | All three | The phone pulls back: the chorus, laughing and waving |
The whip pans are what make it the Gangnam Style video. A version without them looks like any group selfie.
What you need
- One to three photos: one person or one pet per photo, face clearly visible. The video has big close-ups.
- Permission: your own photo, friends who agreed, or your pet. Never a celebrity or a child.
- An AI video tool that takes a reference video: our Gangnam Style AI template, or a model such as Seedance 2.0 with the original clip as a motion reference.
Method 1: the one-photo way (about 10 minutes)
Open the template
Go to the Gangnam Style AI video generator. The original night selfie is already set as the motion reference, with the song ready.
Add the trio
Upload your photo as the main one: you stand in the middle and get both middle solos. Add Friend 2 (the hair flip) and Friend 3 (the glass lobby), or leave them empty and the AI adds two invented friends.
Pick the place and the quality
Gangnam at night, Snowy castle or Tokyo neon. The video is the full 21 seconds, 16:9, with the song laid underneath. It is made in two parts at the same time and joined inside a whip pan, so it costs more than a 10-second template, and the same for one, two or three photos.
Generate and post
You get one 21-second MP4 in about 10 minutes; failed videos are refunded automatically. Post it with #gangnamstyle and label it as AI-generated.
Make the Gangnam Style video with 3 photos
21-second 16:9 video with the trend sound, from 135 credits. No prompt needed.
Open the Gangnam Style generatorWhich place?
| Place | What you get | Best for |
|---|---|---|
| Gangnam at night | The street of the original: office towers, street lamps, Korean flags | The closest remake; it reads instantly as the trend |
| Snowy castle | An old castle courtyard at night in falling snow | Cosplay and costume photos, like the wizard-school versions |
| Tokyo neon | A neon crossing with giant screens | A night-out look with a different city |
Solo, friends or a pet
- One photo: you are the main one, with two solos; two invented friends fill the other spots and get the other solos.
- Two or three photos: your real friends take the right and left spots and get their own solos.
- Pets: a dog or cat in any spot sits up on its hind legs, bobs its head and opens its mouth with the song. It stays the same animal.
- Costumes: upload cosplay photos of yourself and your friends and you keep the outfits; pick Snowy castle for the Hogwarts look.
Photo advice
- A clear face beats a full body. The solos are close-ups; front light, no sunglasses.
- One person per photo. Upload separate photos, never one group shot.
- Dress for a night out. Everyone keeps the clothes from their photo.
- Pets on their own. One animal, face visible, not in someone's arms.
Method 2: make it yourself with an AI video model
The camera is the trend, so a text prompt alone gives you a generic party selfie. You need a model that takes a reference video as well as photos, such as Seedance 2.0. The catch is length: most models stop at 15 seconds and the original runs 21.
Cut the clip in two
Split it inside the whip pan at about 12 seconds, where the frame is pure motion blur. Each part stays under 15 seconds, and the blur hides the join.
Generate both parts with the same photos
Same photos, same order, same setting, 16:9. Tell the model which part it is making and what happens in it.
Join and add the song
Put the parts end to end in an editor, then lay the clip's sound from the first frame so the solos land on the music.
What your prompt has to cover
- Who stands where. Middle, right and left, and the order of the solos.
- The invented friends. With one photo, describe the two invented friends in detail in both parts, or they change between the parts.
- Never copy the original. Say the faces come only from your photos, not from the three women in the clip.
- The whip pans. Ask for the fast handheld whip pans with motion blur between friends.
- Everyone sings. Big mouth movements, singing to the lens.
Common mistakes and how to fix them
- The video stops at 15 seconds. Make it in two parts and join them inside a whip pan.
- A jump at the join. Cut in the middle of the blur, not before or after it.
- The friends change halfway. Use the same photos in both parts, or describe the invented friends the same way.
- The original women appear. Say the faces come only from your photos.
- A group photo for the trio. Upload one photo per person; group photos mix faces.
Label your video as AI-generated where TikTok, Instagram or YouTube ask.
Gangnam Style AI examples
The original night selfie and the AI versions that spread it.
gangnam style in gangnam🇰🇷
КАК СДЕЛАТЬ ТАКОЕ ВИДЕО? 🔥
Hogwarts Gamgamstyle Night...#ai #harrypotter #party
The Lord Of The Rings Gamgamstyle...#ai #lotr #party
قالب CapCut غانغنام ستايل
Make the Gangnam Style video with 3 photos
21-second 16:9 video with the trend sound, from 135 credits. No prompt needed.
Open the Gangnam Style generatorFAQ
Who made the original Gangnam Style selfie?
TikTok user @emiliyyam, who posted it with two friends from a street in Gangnam, Seoul, on 5 October 2026. The song is PSY's "Gangnam Style" from 2012.
Is the Gangnam Style AI trend the horse dance?
No. It copies the night selfie with its close-up solos, not the horse dance from the music video.
How long is the Gangnam Style AI video?
The original runs 21 seconds, and so does our template. Because models make at most 15 seconds at a time, it is made in two parts and joined inside a whip pan.
Can I do it with my dog?
Yes. Upload a photo of your dog or cat in any spot. It sits up on its hind legs, bobs its head and sings along.
Is there a Gangnam Style AI CapCut template?
Yes, but CapCut templates paste faces onto a clip that already exists. An AI generator films the whole selfie again with your faces and your outfits, or your pet.
