12 AI Rap Video Prompts That Actually Work

A good rap video prompt is a shot description, not a mood board. It names the performer, the place, the light and one camera move, then stops. The twelve prompts below follow that structure. They are written for Seedance 2.0 through this studio, and they assume you are uploading a portrait unless the prompt says otherwise.

Use them as starting points. Change the location, the light colour or the action, but keep the identity and guardrail sentences intact. Those are what stop the model from inventing a new person halfway through the clip.

The structure every prompt needs

1. Subject. Who is in the shot, and how many people.
2. Place. A specific location with a surface and a background.
3. Light. Direction, colour and hardness.
4. Camera. One move, one height, one distance.
5. Guardrails. Keep the face and outfit, no text, no watermark, no extra people.

If a prompt is missing one of the five, the model fills the gap with a generic choice. That is where drift comes from.

1. Club stage two-shot

Use the two people in the uploaded photos as the only performers. A small club stage at night, purple and magenta lights cutting through haze, a packed crowd in silhouette at the front. The two rappers stand side by side trading verses, one gesturing to the crowd while the other nods to the beat. One continuous handheld shot at chest height that pushes in from the back of the stage to a tight two-shot. Keep the faces, hair, outfits and body proportions exactly. No extra people, no text, no watermark.

2. Solo booth close-up

Use the person in the uploaded photo as the only performer. A narrow recording booth at night, a condenser microphone on a shock mount, warm tungsten key light from the left and a cold blue rim light behind. The rapper leans into the mic and performs to camera. One slow dolly-in from medium shot to close-up, shallow depth of field. Keep the face, hair and outfit exactly. No headphones swapping, no text, no watermark.

3. Rooftop cypher

Use the person in the uploaded photo as the lead performer. An empty rooftop car park at night, wet concrete reflecting orange sodium lights, a circle of silhouetted figures clapping. The rapper walks into the centre and performs, turning slowly to face the camera. One low-angle tracking shot circling at knee height, ending on a medium close-up. Keep the face, hair and outfit exactly. No fighting, no extra foreground people, no text, no watermark.

4. Retro sidewalk duo

Use the two people in the uploaded photos as the only performers. A sunlit city sidewalk in the mid-1990s, a large silver boombox between them, brick wall and parked cars behind. The pair trade verses and laugh between lines, baggy outfits moving with the beat. One continuous handheld shot with a slight zoom, warm film grain and soft VHS halation. Keep the faces, hair and outfits exactly. No period-inaccurate screens, no text, no watermark.

5. Drill rooftop, cold and wide

Use the person in the uploaded photo as the only performer. A wide empty rooftop at blue hour, cold overcast light, distant tower blocks out of focus. The performer stands still with a hood up, then turns to camera and steps forward once. A slow low-angle push-in from a wide shot to a medium shot. Keep the face and outfit exactly. No weapons, no crowd, no text, no watermark.

6. Golden-era basement

Use the person in the uploaded photo as the only performer. A low basement with exposed brick, a single warm tungsten bulb overhead, a turntable and stacked records behind. The rapper performs to camera with one hand marking the beat. A locked-off medium shot with a very slow zoom. Warm film grain, soft contrast. Keep the face and outfit exactly. No readable labels, no text, no watermark.

7. Tunnel walk

Use the person in the uploaded photo as the only performer. A concrete pedestrian tunnel at night, repeating overhead lights, wet floor reflecting the performer. The rapper walks toward the camera while performing, hands moving with the beat. One steady backward-tracking shot at chest height. Keep the face, hair and outfit exactly. No extra people, no graffiti text, no watermark.

8. Stage silhouette reveal

Use the person in the uploaded photo as the only performer. A dark stage with a single white spotlight from behind, haze in the air. The performer starts as a silhouette and steps forward into the light, revealing the face. One slow push-in from wide to medium. Keep the face and outfit exactly. No band, no crowd, no text, no watermark.

9. Crowd-surf energy, no crowd faces

Use the person in the uploaded photo as the only performer. A packed club seen from the stage, hands raised in silhouette below, magenta and blue lights sweeping across haze. The rapper leans over the front edge and performs to the crowd, then looks down the lens. One handheld shot at chest height with a slight push-in. Keep the face and outfit exactly. No individual crowd faces, no text, no watermark.

10. Night drive backdrop

Use the person in the uploaded photo as the only performer. The back seat of a moving car at night, city lights streaking past the window, cool blue interior light on the face. The rapper performs to camera, one hand on the seat. One locked-off shot with slight handheld float. Keep the face and outfit exactly. No visible driver, no dashboard text, no watermark.

11. Lyric-video visual bed

Use the person in the uploaded photo as the only performer. A dim room with a single practical lamp, deep shadows, the performer framed slightly to the right with empty wall on the left. The rapper performs quietly to camera, minimal movement. A locked-off medium shot with a slow drift. Keep the face and outfit exactly. No text, no watermark.

12. Label showcase

Use the two people in the uploaded photos as the only performers. A bright photo studio with a seamless grey backdrop, soft even light. The pair stand relaxed, one gesturing toward the other, then both look to camera. One slow dolly-in from medium to medium close. Keep the faces and outfits exactly. No logos, no text, no watermark.

How to adapt a prompt without breaking it

Change one block at a time. If you swap the location, keep the light and camera sentences identical so you can tell whether the location change helped. If you change the camera, keep the location and light fixed.

Do not add a second camera move to a five-second clip. Do not add a crowd and a car and a wardrobe change in the same render. Every extra element is another thing that can drift, and the model does not prioritise the way a human crew would.

The guardrail sentence is not optional

The last two sentences — keep the face and outfit, no text, no watermark, no extra people — are doing most of the consistency work. They are boring, and they are the reason a clip is usable. If you rewrite a prompt from scratch, copy those sentences across.

For more on the model itself, read Seedance 2 rap videos. For the full workflow around a prompt, see how to make an AI rap video.