The short version
The trend started with a clip of a young man in a black STAFF polo dancing at a quinceañera, known as the BBY WOW guy. The AI version keeps that clip and swaps him for someone else. Nobody dances: the model generates new video in which a different subject performs his walk and moves.
People ask how the videos are made because it looks like you need to find the dance and learn it. You don't. The original clip is already built into the tool as the motion and scene reference.
What goes in
Two inputs. First, your photo: one person, character or pet, ideally the whole body with the face clear. Second, the original clip, cut to vertical and with its audio removed, which supplies the room and the motion.
The instruction we give the model is plain: use the face, hair, body, skin and outfit from the photo; keep the party hall, the purple and green lights, the guests behind, the camera movement and his exact dance moves, walk and timing; change only the dancer.
What changes and what stays
From your photo comes who is dancing: face, hair, body, outfit. From the original clip come where and how: the room, the lighting, the other people, the camera path and the choreography.
- Changes: the dancer's face, hair, body and clothes
- Stays: the quinceañera hall and lighting
- Stays: the guests in the background and the camera movement
- Stays: the dance moves and their timing
- Not included: the song. The file has an AI-made beat because we don't have the rights to BBY WOW
What happens to a pet
A pet takes the dancer's place in the scene and copies his moves and position instead of dancing the way an animal would. In our cat example, the cat keeps its own face, fur and colouring and is not turned into a person. We have only tested a cat so far, not a dog, so we don't promise the same result for every animal.
Where it goes wrong
AI video generates frames rather than filming them, so odd frames happen: a hand that bends, a face that shifts slightly as the head turns. Most trouble starts with the photo: several people in it, a face that is tiny or covered, or heavy filters. Our guide to photos that work covers the fixes.
The finished video is about 9 seconds long, 720p (720×1280) and vertical. If it does not look like the person or pet in your photo, email us and we refund that video.
What happens to your photo
Your photo is sent to the AI video model that produces the video, and hosting, database, payment and email providers handle the data needed to run the service. Uploaded photos are deleted automatically 30 days after upload, and we don't use your content to train AI models. The privacy policy has the full list.
Ready to try it? Upload one photo on the home page. Pricing is on the pricing page.