Multi-character Gangnam Style: two or three photos in one AI dance video
The photo-to-dance trend is shifting. Solo clips — one person, one meme dance — were the first wave. The clips that stand out now are groups: two friends, a couple, a trio of characters doing the horse-riding dance side by side in one shot. This site's generator now supports that directly: upload one to three reference photos, one person or character per photo, and the request is sent as a multi-reference job.
What actually happens with multiple photos
Single-photo requests use an image-to-video model: one reference drives the whole clip. Two- and three-photo requests switch to a reference-to-video model. Each photo is kept as its own reference, and the prompt addresses them in order — the character from the first photo, the character from the second photo, and so on. The model is asked to place them side by side, keep each character's face, outfit and proportions from their own photo, and give each a slightly different dance variation so the row reads as a party rather than a mirror.
A verified example, not a promise
On October 22, 2026 we generated a duo test with two studio-style character photos: a woman with a red bob haircut in a yellow jacket, and a man with curly black hair, glasses and a green hoodie. The 5-second request finished in about two minutes as a 1280×720 MP4. Both characters appeared in frame from the first beat, each kept their own hair, outfit and proportions, and both danced with their own variation. You can watch the unedited clip in the showcase on the home page. One test is an observation, not a guarantee: identity drift, merging and missing characters can still happen, especially with crowded or similar-looking references.
For the current three-photo steps, use our dedicated trio tutorial and trio generator page.
How to make a duo or trio clip
- Prepare one photo per character. JPG, PNG or WEBP, facing the camera, bodies visible, clear light. Distinct silhouettes and different colours help the model keep the characters apart.
- Upload two or three photos in the generator and pick a length. Groups can request up to 19 seconds; solo requests cap at 15. Add a short character note if you want to steer the scene.
- Review every character through the whole clip before posting. Check faces, outfits and hands, and confirm no character merged or vanished. Download links last one hour; the saved task in the same browser can refresh the link.
What to watch out for
- Multi-character clips have more failure modes than solo ones. Request a short clip first, review it, then try longer ones.
- Use photos you have permission to upload, with the consent of anyone shown. Generation does not grant rights to a person, character or song.
- Follow your platform's AI-content labeling rules. Realistic AI-generated people dancing should be labeled as AI-generated.
Try it during early access
Request trial availability or package details from Billycui50@gmail.com. Online payment is not connected. The default preview is 5 seconds; longer access is approved separately. Bring one photo for a solo clip, or two or three for a group.
Read the generation guide, browse prompt examples, or see a group concept in Spider-Man trio planning.
Other character concepts: Gandalf · Harry Potter · Kara Sevda