Carpool Karaoke AI Prompt: A Seat-by-Seat Template

This Carpool Karaoke AI prompt recreates the Walk It Like I Talk It car scene in AI video models that accept several reference images and a reference audio track. It casts each seat from its own image and follows the shot list we measured on the viral clip. It’s the same structure our generator uses — if you don’t want to tune it yourself, the generator does it from your photos.

What to feed the model

1. Image 1 – the set: an empty SUV interior seen from the dashboard, daylight, beige leather seats, a sunroof.
2. Images 2–5 – the riders: one photo per seat, in this order: driver, front passenger, back left, back right. One face per image.
3. Audio 1 – the hook: the audio of the trend, starting at the first “Walk it like I talk it”.
4. Settings: 15 seconds, 9:16 or 16:9, audio generation on.

The Carpool Karaoke AI prompt

Vertical 9:16 video in a car. Shot like a TV carpool karaoke segment in daylight: a beige-leather SUV with a sunroof, a sunny suburban street outside, hard cuts between angles, no cameras or devices visible on the dashboard. Image 1 is the set: keep this car interior in every shot.
The driver is the subject of image 2, on the right side of the frame behind the steering wheel. The front passenger, who leads the song, is the subject of image 3, on the left side of the frame. The subject of image 4 sits in the back left and the subject of image 5 in the back right.
Keep every face, hairstyle, fur and outfit exactly as in its own image and never swap or blend faces between seats.
0-3s: close-up of the two back-seat passengers bobbing their heads and rapping. 3-4.3s: side shot of the driver at the wheel, one back-seat passenger visible behind. 4.3-5.2s: close-up of the front passenger leaning into the lens with a hand gesture. 5.2-8.3s: wide windshield shot of all four riders. 8.3-10.5s: close-up of the front passenger rapping straight into the lens. 10.5-11.8s: close-up of the driver, eyes wide, mouth open on the beat. 11.8-12.7s: close-up of the front passenger. 12.7-14s: wide windshield shot, everyone bouncing and pointing at the lens. 14-15s: close-up of the driver.
Audio 1 is the soundtrack of this video and plays unchanged from the first frame. Everyone raps along to the vocals in Audio 1, lip-syncing every word, and every move lands on the beat.
Realistic. No text, no logos, no subtitles.

With a pet in the car, add: If a rider is an animal, it sits upright in its seat like a person, bobs its head to the beat and opens its mouth in time with the vocals.

The reference-video version of the Carpool Karaoke AI prompt

If your tool can follow a reference video (like Higgsfield Genjutsu), use a clip with the same cuts as the reference and describe only the swap:

Video1 is the reference for camera, cuts, timing and every movement. In every shot, replace the driver in Video1 with the subject of Image1; replace the front passenger in Video1 with the subject of Image2; replace the back-left passenger in Video1 with the subject of Image3; replace the back-right passenger in Video1 with the subject of Image4. Keep each face, hair, fur and outfit from its image and never blend faces. Keep everything else from Video1 unchanged: the car, the daylight, the camera positions, the cuts, the other riders and the soundtrack. Realistic. No text, no logos, no subtitles.

Tips for any Carpool Karaoke AI prompt

• Don’t quote lyrics in the prompt. Models improvise when they see lyrics, and the timing drifts. Let the reference audio carry the words.
• Write “image 2”, not “@Image2”, in models that number references in order; the @ form can change the camera to a selfie view.
• One face per reference image. Group photos are the main cause of blended faces.
• Name the seat and the side of the frame for each rider. The driver sits on the right of the frame in a US car seen from the windshield.
• Don’t name real people. Describe seats, not celebrities.
• Expect the model to re-sing the hook with a short lead-in. If you want the exact original audio, replace the track in an editor and line it up with the mouths. Our generator measures this lead-in on every video and does it for you.

Skip the Carpool Karaoke AI prompt: ready-made instead

Carpool Karaoke AI runs this setup from 1–4 photos, adds the original hook lined up, and returns a 15-second HD MP4 for $9.90. See how to make a Carpool Karaoke AI video for the step-by-step.