Three photos. One dance idea.
Create a Gangnam Style AI Trio Video from 3 Photos
Prepare one clear photo for each of three people or original characters, then try a Gangnam Style-inspired group request. Identity and choreography can vary, so inspect every dancer before sharing.
Your photos. Their dance floor.
Add one photo per person, in the order you want to request. Use all three for a trio. Placement and likeness are model-guided and can vary.
JPG, PNG or WEBP · up to 10 MB per photo · one person per photo · clear body and face · use images you have permission to upload
The length options are requests to the video model. Final duration, dimensions and dance accuracy can vary. This tool does not automatically crop every image to 9:16.
Prepare the people, then try the movement
Use a separate image for each dancer. A clear face helps with identity, while visible shoulders, arms and legs give the model more information for the horse-riding movement. Start with similar lighting and scale across all three references. Tiny faces, heavy sunglasses and overlapping bodies make the request harder.
01 / Three references
One person per photo. JPG, PNG or WEBP. Keep each face visible.
02 / One group request
Ask for a stable camera and a simple left, center and right arrangement.
03 / A result to inspect
Check all three people throughout the clip before saving and posting.
How the trio workflow works
Open the main generator, select all three images, choose an available duration and enter the access details arranged for your trial. The built-in prompt requests a Gangnam Style-inspired horse-riding dance. Optional notes can identify clothing or request positions; they cannot guarantee exact placement or facial consistency.
When a task completes, preview the MP4. A longer duration creates more opportunities for a face or costume to drift. If an identity becomes unstable, a shorter attempt and simpler references may be easier to assess. The friends-dance guide explains how to crop a group photo and check for swaps.
What the real preview tells you
The embedded output shows two photo references becoming a two-character dance clip. Its duration is 5.042 seconds and its dimensions are 1280 × 720. It gives you a useful example of motion and composition. It does not establish three-person consistency, exact choreography, a 19-second result or a standard output size for every request.
Access, output and sharing
Payments are not connected. Contact Billycui50@gmail.com to ask about trial access and introductory discounts. An allowance, duration and resolution must be confirmed for your trial. See the proposed plans before choosing a longer or higher-resolution request.
Generated video does not include a music licence. Save the MP4 and add music available for your use on the destination platform. For a longer edit, follow the 19-second planning guide; for audio, see the TikTok music workflow.
Frequently asked questions
Can I upload three separate photos?
Yes. The generator accepts up to three JPG, PNG or WEBP reference images in one request. Use one person or original character per image and check the previews before submitting.
Will all three faces stay the same?
There is no guarantee. Multi-person generation can blend faces, swap clothing, move people or omit a character. Review every dancer at the beginning, middle and end of the result.
Can I request a 19-second trio?
The multi-photo workflow supports requests up to 19 seconds after longer-request access is confirmed and configured by the site owner. The standard trial starts at five seconds. A published three-person 19-second test is not yet available, and an exact final file length is not guaranteed.
Is the video on this page a trio sample?
No. It is a real two-character result from two photos, lasting 5.042 seconds. It is clearly shown as a duo reference, not proof of a three-person or 19-second output.
Can I try the trio generator for free?
Email Billycui50@gmail.com to ask for trial access and current introductory discounts. Payments are not connected, and a free allowance, duration or output resolution is not automatically assigned.