ADÉLA duo
Two photos; follow the complete night-plaza dance, with fixed left and right identities.
ADÉLA Dance Challenge: Nicole Kidman AI Dance Tutorial
Make a two-person ADÉLA dance video from two identity photos and a built-in motion reference. Assign left and right dancers, review the credit quote and follow your task.
Upload one photo for each dancer. The built-in vertical reference guides the night-plaza movement; your photos define the two appearances. This workflow requests 15 seconds and silent output. The preview is the reference, not a generated result.
Try this effect →Create your video
Two photos; follow the complete night-plaza dance, with fixed left and right identities.
Checking your sign-in…
JPG, PNG or WebP, up to 10 MB. Use one clear subject per photo.
Built-in motion reference · rounded duration used for billing · 15 seconds
MiniMax H3 · 768P · Reference-video seconds and output seconds are included in the quote. Images and audio use the existing pricing policy.
Silent video only. The built-in clip has no song, and this template does not accept an audio reference.
Read the English motion prompt
Edit @Video1. @Image1 replaces ONLY the dancer starting on the LEFT; @Image2 replaces ONLY the dancer starting on the RIGHT. Each image is an independent full identity reference, never a first or last frame. Replace face, hairstyle, body appearance, clothing and accessories from each assigned photo. Track the same person throughout movement and turns, without exchanging faces or blending identities. Preserve the night plaza, lighting, low stationary camera, framing, original action order, rhythm and full finish from @Video1. Do not force dancers to face the camera during a reference turn. Preserve coherent hands and feet. Keep the reference complete and hold the final pose only for the sub-second rounding remainder. Silent output; do not recreate any existing song or voice.
Extra direction and the sound preference are added at submission. Model guidance does not guarantee frame-by-frame matching.
Estimated credits：23 credits
Sign in to createADÉLA Dance Challenge: Nicole Kidman AI Dance Tutorial · Prepare the inputs
Use one clear photo of a consenting adult per slot, ideally showing their full outfit. Photo A replaces the person starting on the left; B replaces the person starting on the right. Each is an identity reference, not a first or last frame. The reference keeps the low camera and night setting, so a photo’s background does not select a new location.
Choose a variation
How to make the video
Choose MiniMax H3 or Seedance 2.5 and a supported quality. Add both photos, check their labels, and edit the extra direction if needed. Wait for the server quote, confirm your available credits and generate once. Keep the task number; refreshing queries that task rather than submitting another video.
Review the result
Watch the full video, including raised arms, turns and the finish. Check that each face, hairstyle and outfit stays with the same dancer when positions change. Examine hands, feet and separation between bodies. A successful task status does not certify those visual details, and matching every frame is not promised.
If generation cannot start, check both photos, the selected quality and the quote. For an uncertain submission, refresh the existing task or open My Creations before trying again. A failed task follows the site’s credit refund process. Do not resubmit a task merely because it is still queued.
Common questions
How are the two photos assigned?
A is the dancer starting on the left and B is the dancer starting on the right. The complete face, hair, body appearance and clothing come from each assigned photo. Roles stay attached to identities through the turn.
What are the length and audio settings?
The silent reference is about 14.30 seconds, rounded to 15 whole seconds for the request and reference quote. The final fraction may hold the last pose. This template does not include the song or accept an audio upload.
How much does it cost, and has the output been tested?
The server quote uses the chosen model, quality, 15-second output and 15-second reference. Generation uses site credits; failed tasks follow the refund process. This template has passed offline request checks, but no paid output test was run for this release.
The generated motion, likeness and soundtrack need to be reviewed before downloading.
Read this page as Markdown