Practical guide / three-photo workflow
How to Make the Gangnam Style AI Trio Video
Prepare three clear photos, one per person, then upload them together in the generator. This guide covers the decisions that matter: useful references, simple position notes, duration requests and a careful check of every face in the result.
What is the Gangnam Style AI trio workflow?
It is a photo-to-video request that asks three referenced people or characters to appear together in a horse-riding dance inspired by Gangnam Style. The model creates motion from the images and prompt; you are not pasting three faces onto a fixed performance. That also means choreography, face stability and timing can vary.
What you need before starting
- Three separate JPG, PNG or WEBP references with one intended person or original character per image.
- Permission from the people shown and permission to use the source images.
- Confirmed access: email Billycui50@gmail.com to ask about a trial and current discounts. Payments are not connected.
- A clear idea of the intended framing and whether you want a short test or a longer group request.
Step-by-step: make three people dance with AI
Step 01Prepare one clear photo per person
Choose three JPG, PNG or WEBP images, with one person or original character in each. Prefer a visible face, torso and legs, similar lighting and a similar camera angle. Get permission from everyone pictured. Crop a group photo into separate references if each person remains clear.
Step 02Upload all three references in the generator
Open the generator and select the three files together. Check each preview before continuing. Use the character notes to describe visible clothing and the intended left, center and right arrangement. Placement is a request and may change in the output.
Step 03Choose an available length and request a clip
Start with the five-second trial request unless different access has been confirmed. The service supports multi-photo requests up to 19 seconds and solo requests up to 15 seconds, but longer requests need access configured by the site owner. Arrange trial details by email. Submit once, then use the saved-task check if you return to the page.
Step 04Review the three identities and save the MP4
Watch the whole clip for missing people, blended faces, swapped clothing and unstable hands. Download the completed MP4 while its link is active; the generator can refresh a saved task link while the output remains available. Add permitted music separately before sharing.
How to keep all three faces easier to recognize
Use roughly similar face sizes and camera angles. A front-facing image with open eyes is easier to compare with the result than a profile hidden behind hair. If one person is far away in a group picture, ask for a separate source photo instead of enlarging a tiny crop.
Describe visual anchors rather than asking the model to remember names. For example, “red jacket on the left, blue shirt in the center, green sweater on the right” is more concrete than three first names. Treat that order as a request: it can still change. The friends-dance guide goes deeper into crops, placement and identity swaps.
Fix the specific problem you can see
| What you see | What to check next |
|---|---|
| Two faces blend together | Use clearer separate references, reduce overlap and simplify the requested scene. |
| One dancer disappears | Check all three uploads are present and ask for a wider stable composition. |
| Clothing or position swaps | Use distinct visible outfit details in notes. Avoid claiming positions will be fixed. |
| Movement starts late | Review the original first. Trim a pause in an editor if that suits the clip; do not describe an edit as new generated motion. |
| Status or download link stops working | Check the saved task in the same browser before creating a duplicate request. Refresh a completed link while the output remains available. |
Download and share a result you have checked
Save the MP4 and watch it locally before editing. Inspect the real duration and dimensions; a selected length is a request, and the service does not guarantee the same framing for every input. If you need a vertical post, crop carefully so the outer dancers remain in frame.
Add a sound available for your account and intended use, then preview the finished edit with the people pictured. The TikTok music tutorial covers the sound stage. For a longer composition, follow the 19-second planning guide.
Frequently asked questions
Can I start from one group photo?
You can crop it into separate references if every person remains clear. The multi-photo workflow expects one intended person or character per image; tiny or overlapping faces are a weak starting point.
Does uploading three photos guarantee a successful trio?
No. It makes a three-reference group request possible, but the model can omit a person, blend faces or change clothing. Inspect the completed output before sharing.
Should my first request be 19 seconds?
A short request is useful for checking motion and identity if your access permits it. Multi-photo requests accept up to 19 seconds, but the site has not published a verified three-person 19-second result.
Do I need a CapCut template?
No template is required for this generator’s image-to-video workflow. You can use an editor afterward to trim, crop, add permitted music and prepare the post.