Final step: the location + video Generate an empty street first, so the AI has a fixed place to put both characters. Street prompt: Empty Paris café terrace at dusk, rattan bistro chairs and small round tables, grey awnings, glowing wall lamps, trees and wide grey stone paving. No people. iPhone photo. Video prompt (upload both sheets + the street, 1080p, 9:16, audio on): Selfie video on this café terrace. The Tudor king screams into the front camera, clutching his head, while the man in the fedora dances behind him. The phone switches hands and they swap places. Then both dance side by side in sync, filmed low from the pavement. Shaky handheld phone footage, hip-hop beat. That's it. 2 characters, 1 street, 1 prompt.

Step 4: The video prompt ([model name]) Upload: Image 1 = first frame, Image 2 = location sheet, Image 3 = character sheet Settings: 1080p, 9:16, 10s, audio on The secret: don't describe a scene. Describe the PHONE filming it. Prompt: === REFERENCES === Image 1 → opening frame. Image 2 → venue. Image 3 → character. Keep his bowl cut, handlebar moustache, outfit and mismatched socks exactly the same throughout. He grins, raises his eyebrows and pulls playful faces while he dances. === CAMERA === Handheld iPhone held up by someone in the crowd, at slightly below head height. Small natural wobbles, slight reframing, occasional hands and phones crossing the lens. No cuts, no zoom, no smooth gimbal movement. === ACTION === He dances with big comedic energy: claps above his head, fires finger guns at the crowd, blows a kiss, then breaks into a wild sideways lunge and shimmy. Every pose is exaggerated and theatrical. === CROWD === Starts out filming calmly. By the end they're cheering, punching the air and waving. === AUDIO === Bouncy disco-pop track through party speakers, with the crowd whooping and cheering. Recorded on the phone mic.