Carpool Karaoke AI Prompt Guide
Seat order, photo tips, and the fixed Wan 3.0 reference prompt for four-person Carpool Karaoke AI videos.
Last updated: 2026-10-13
On this site you do not need to paste a prompt. Upload four seat photos (driver, front passenger, rear left, rear right) and we apply the fixed Carpool Karaoke windshield-camera reference plus a stable English Wan prompt.
Open homepage generatorEach card explains casting or photo prep and shows the same English Wan R2V seat prompt used conceptually on this site. On the homepage you do not paste it — we apply it with the fixed reference video.
Seat order (casting plan)
Image1–4 map to fixed seats
- Image1 = front passenger, Image2 = driver, Image3 = rear-left, Image4 = rear-right
- Put your most expressive friend in the driver seat when you want big energy
- If you only have three people, duplicate someone into a rear seat — do not leave a slot empty
Full English Wan R2V prompt
Use Video 1 as the exact motion and edit reference. Keep the same car cabin, seating layout, camera angles, cuts, timing, gestures, expressions, and lip-sync. Video 1 controls camera, cuts, motion, and performance. The four reference images control only identity and wardrobe. Fixed seat map for every frame and every cut: - Image 1 → front passenger - Image 2 → driver - Image 3 → rear-left - Image 4 → rear-right Replace all four original people in Video 1 with these four subjects. Preserve each subject’s exact face, hair, skin tone, and outfit from their image throughout the whole clip. Do not swap seats. Do not blend or mix features between subjects. Do not let any original Video 1 face return after a cut or near the end. Keep the original car and framing. Do not redesign the cabin, invent new actions, add people, or smooth Video 1 cuts into drifts.
Photo tips for all four seats
One clear person per upload
- Front or three-quarter face, eyes visible, even lighting, minimal blur
- Never put a group shot in one slot — two faces confuse casting
- Avoid heavy sunglasses/masks and extreme crops that cut the jaw or forehead
Full English Wan R2V prompt
Use Video 1 as the exact motion and edit reference. Keep the same car cabin, seating layout, camera angles, cuts, timing, gestures, expressions, and lip-sync. Video 1 controls camera, cuts, motion, and performance. The four reference images control only identity and wardrobe. Fixed seat map for every frame and every cut: - Image 1 → front passenger - Image 2 → driver - Image 3 → rear-left - Image 4 → rear-right Replace all four original people in Video 1 with these four subjects. Preserve each subject’s exact face, hair, skin tone, and outfit from their image throughout the whole clip. Do not swap seats. Do not blend or mix features between subjects. Do not let any original Video 1 face return after a cut or near the end. Keep the original car and framing. Do not redesign the cabin, invent new actions, add people, or smooth Video 1 cuts into drifts.
Wan R2V seat prompt
Reference video + four identity images
- Match the reference video’s camera, seats, timing, and motion
- Keep each face identity with natural lip sync; no extra people, text, or watermark
- Homepage output is ~15 seconds at 720p with aspect ratio you choose
Full English Wan R2V prompt
Use Video 1 as the exact motion and edit reference. Keep the same car cabin, seating layout, camera angles, cuts, timing, gestures, expressions, and lip-sync. Video 1 controls camera, cuts, motion, and performance. The four reference images control only identity and wardrobe. Fixed seat map for every frame and every cut: - Image 1 → front passenger - Image 2 → driver - Image 3 → rear-left - Image 4 → rear-right Replace all four original people in Video 1 with these four subjects. Preserve each subject’s exact face, hair, skin tone, and outfit from their image throughout the whole clip. Do not swap seats. Do not blend or mix features between subjects. Do not let any original Video 1 face return after a cut or near the end. Keep the original car and framing. Do not redesign the cabin, invent new actions, add people, or smooth Video 1 cuts into drifts.