HypeStormWhat is the Raindance AI trend?
The Raindance AI trend puts new people into music-video scenes inspired by Dave and Tems’ "Raindance". The common seaside version follows a fixed sequence: one performer alone on a sunlit pier, the other singing to camera by the water, the two together on the sand, and an ending on the beach at dusk. Creators recast both performers with photos of themselves, friends, or characters — no filming, no dancing, no camera.
Like the hotel-lobby and zombie trends before it, the video follows a fixed script with two fixed roles. That structure is exactly why AI nails it — the model only has to hit each shot of the script and keep two identities straight.
Two photos, two fixed roles
The trend’s one rule: assign an image to each performer and keep that mapping in every shot. Image 1 replaces the one on the pier; Image 2 replaces the singer. Either role can be filled by a person of any gender.
• Use one clear, front-facing photo per person — good light, no sunglasses.
• Full-body photos beat portraits: the model carries hair, body proportions and the complete outfit.
• After rendering, check every singing close-up — a performer reverting to the original face is the classic failure mode; re-run that segment with the same image.
The 28-second sequence
Shot-for-shot versions run just under half a minute, in four beats cut to the song:
• 0–4s — The wait: one of you sits alone at the end of a sunlit wooden pier while the other walks up behind.
• 4–18s — The singer: close-ups by the water and moments on the pier, with the two of you side by side on the sand.
• 18–25s — Together: a playful photo shoot on the pier, then the singer’s last close-up.
• 25–28s — The ending: side by side on the beach at dusk, then one last close moment on the sand.
Which tools can make it?
Two routes. Reference-clip workflows (Viggle-style character replacement, and editors like Media.io’s raindance generator) swap both performers into the original sequence — cuts, timing and camera stay recognizable, which is what makes the trend land.
From-scratch generators (Higgsfield, Gemini/Veo, Kling) animate each beat from your photos — more control, but you’ll write and tune a prompt per shot, cut eleven clips to the song’s timing, and expect several retries per shot to keep faces locked. That trial-and-error cost is the hidden price of the manual route: prompt tuning alone typically takes an evening.
Or make it here in one step
HypeStorm runs the whole sequence for you: upload two photos, pick a resolution, and the generator renders all eleven shots of the 28-second seaside duet with both performers recast and every cut timed to the song — no prompt engineering, no stitching. Try a free 4-second preview first.
Posting it right
Add "Raindance" from the platform’s own music library when you post — not a downloaded copy — so your video joins the trend’s sound page, where the first-page views live. The video renders silent on purpose: the track you add is the trend’s audio. Leave the sound starting at 0:00: the video is cut shot for shot to the song, so the singing close-ups land where people expect them with no manual syncing. HypeStorm renders vertical 9:16 out of the box; if you made yours elsewhere, crop to 9:16 where it preserves the action.