Short social formats win attention by being instantly recognisable and easy to remix. The “Alors on danse” AI trend does both: it re-animates a compact 2010 sketch—one person on a small keyboard, another hyping the performance—by swapping in new faces with motion-transfer tools. For social teams and creators, that makes the meme a low-friction opportunity to test character-driven hooks and rapid creative iterations.

Why this matters for social video

The format’s value is practical. It’s a controlled, two-shot scene with consistent framing and audio, which reduces the complexity of motion-transfer generation compared with freeform lip-sync or full-scene synthesis. That translates to faster iterations, lower production cost and predictable creative constraints—useful for brands testing audience response, creators producing bite-size content, and social teams experimenting with personality or character pairings.

Step‑by‑step: make your own Alors on danse AI video

The workflow below reflects how creators are using dedicated AI generators and motion-transfer tools to remake the sketch. The example mentions a Kapwing-style custom generator and a Kai workflow, but the steps apply to similar tools that support two-character motion transfer.

1. Pick the two roles and the pairing

Keep the original dynamic: one person plays the keyboard (focused, driving the music) and the other is the expressive hype partner (gestures, reactions, crowding the frame). For viral potential, choose a pairing that’s either immediately recognisable or humorously incongruous—two public figures, a creator and a fandom character, or an unexpected fictional duo.

2. Prepare clean reference images

Motion-transfer quality depends on reference photos. Use clear front-facing or three-quarter portraits with good lighting and visible facial features. If the characters need to turn or show a range of expressions, assemble a simple character reference sheet so the model can maintain consistent identity during head turns and expression changes. Avoid heavy occlusion (sunglasses, extreme angles) and prefer higher-resolution images to reduce identity drift.

3. Use a custom generator and assign roles

Open the meme-specific generator (the Kapwing example provides a ready workflow). Upload the two reference images and explicitly assign each to the correct role where possible: Image 1 = keyboard player; Image 2 = animated partner. Many custom workflows already include the original “Alors on danse” audio, so you may not need to add music separately.

4. Generate, review and iterate

Submit the generation and review the output carefully. Two-person generations face unique challenges because the model must keep two identities distinct while animating interactions. Watch for three common problems and fixes:

  • Faces look too similar: Use more visually distinct reference photos and instruct the generator to preserve separate identities.
  • One character becomes unrecognizable during movement: Replace that character’s photo with a clearer, higher-resolution portrait showing more facial structure.
  • Characters appear in swapped roles: Ensure images are assigned to the intended positions and add a short role instruction if the workflow accepts text direction.

5. Make finishing edits for social

Export the clean generation into your editor and prepare platform-ready cuts: crop to 9:16 for TikTok and Instagram Reels while keeping both characters and the keyboard visible; add a caption that frames the joke (lead with the surprise duo); or create a reaction edit pairing the AI clip with real footage. Consider compiling multiple pairings into a short montage to A/B which duo performs best.

Practical considerations and risk checkpoints

Generative face-and-voice content offers creative upside and operational risk. Before publishing, run these checks:

  • Consent and likeness rights: Obtain consent from any real person whose likeness you use. With celebrities or public figures, evaluate brand and legal risk—platform policies and local laws vary on synthetic content.
  • Copyright and music: The trend uses Stromae’s “Alors on danse” (2009). If your generator includes the original audio, verify licensing or platform reuse policies in your market.
  • Platform rules: Short-form platforms have different policies on synthetic media and deceptive content. Label AI-generated clips where required and follow community guidelines.
  • Brand safety: Avoid pairings that reference sensitive conflicts or explicitly copyrighted characters without clearance. Run risky concepts through a small internal review before wide release.

Distribution and measurement tips

Move fast and learn: produce multiple pairings, A/B captions and thumbnails, and track engagement metrics—view-through, replays, shares and comments. Because the format is recognisable, early performance will hinge on the pairing and the hook: lead with a caption that names the surprise duo or teases the punchline, then iterate on the combinations that generate conversation.

Context to remember: the trend remixes a 2010 Made in Jamel sketch starring Stromae and Jamel Debbouze and re-emerged in 2026 after AI remakes proliferated online. One widely shared recreation used AI-generated likenesses of Sam Altman and Dario Amodei as a topical example of competing AI figures—an illustration of how topical pairings accelerate spread.

What to watch next

Monitor platform policy updates and early audience signals. Prioritise consent and clear labeling, and be prepared to pause or adjust a campaign if it raises legal, platform or reputational flags. Practically: start with small tests, measure which pairings spark conversation, and scale only those that pass your legal and brand checks.