Choose the right source format
The trend is a two-person workstation sketch, not a full music video. Keep that distinction in mind before you brief a general video model.
Full walkthrough
How to make an Alors on Danse AI video
Start with the real studio format, upload two photos, map the roles correctly, and let the prepared template carry the camera, movement, and edit.
The origin is important: this trend is associated with the Made in Jamel studio sketch, not an official Stromae music video. That distinction helps when you write a brief for a general model that keeps trying to make a music video instead of a short two-person sketch.
Create VideoSix steps from two photos to a finished clip
The order matters more than keyword stuffing. Each step removes a failure mode that would otherwise show up in the render.
Prepare the workstation photo
Use one clear portrait for the person who should operate the keyboard and laptop. Keep the face large and the expression neutral.
Prepare the visitor photo
Use one clear portrait for the person who should react beside the desk. Glasses, hair, and facial hair should stay visible.
Assign the roles in order
The workstation host goes first and the visitor goes second. This order prevents the model from cross-swapping identities.
Choose the output frame
Use widescreen for embedded posts, 9:16 for short-form feeds, or square for mixed timelines. The action stays the same.
Render, inspect, and download
Check the credits, submit the task, inspect the face consistency, and download the version that fits the channel.
What the photo pair is doing
The first photo controls the workstation host. The model uses it for the person who operates the keyboard and laptop, stands near the center of the interaction, and anchors the opening wide shot. The second photo controls the visitor, who reacts beside the workstation and provides the quicker facial expressions.
Because the source clip already contains two performers, the strongest results come from keeping the roles separate from the start. You are not asking the model to invent a new scene. You are asking it to preserve a known interaction while changing two identities. That is why the prepared website workflow is more predictable than writing a fresh prompt for a general video tool.
If you need a reusable text prompt, the prompt library includes Seedance, Gemini with Flow, Genjutsu, and a single-image fallback. If you want the shortest route, use the generator at the top of the home page.
Common mistakes
A group photo in one slot
The model cannot reliably choose which person to swap. Crop each person separately before uploading.
The roles in reverse order
The host and visitor perform different actions. Reversing the slots creates identity drift around the keyboard and close-ups.
Heavy beauty filters
Filters remove the small face details that keep the person recognisable across cuts, glasses, and profile angles.
Ready to render?
Use the production generator to upload the two photos, set the format, and check the credit cost before you submit.
Create Video