core duo

AI Rap Video Duo Prompt (Two Photos)

The flagship AI rap video prompt: two separate portraits become a left/right duo performance — traded bars, fixed sides, a hanging mic between them, and a clean 8-12 second loop. Every other template in this wiki is a variation of this structure.

When to use this template

Use it when you have two people, two clear portrait photos, and permission from both — a friendship, a partnership, a birthday duo, a "we actually did it" post. If both photos are of you, switch to the self-duo template; if the energy is romantic rather than hype, the couple template reframes it.

Hook & verse structure

The duo format lives on call-and-response. Before prompting, write four short lines: the left performer owns lines 1 and 3 (the setup), the right performer owns lines 2 and 4 (the answer). That alternation is what makes the clip feel like a real trade instead of two people talking over each other. Keep each line under 10 syllables — image-to-video models hold lip-sync best on short phrases, and the original lyrics template generates them to fit.

Copy-ready main prompt

Create an AI rap video duo performance from two separate portrait references.
Left performer = photo 1, right performer = photo 2.
Keep both faces recognizable, front-lit, and assigned to fixed sides for the whole clip — the left performer never crosses to the right, and vice versa.
Stage: original studio booth, warm orange key light, a single hanging microphone centered between them, dark background with soft rim light.
Wardrobe: keep each performer's outfit exactly as shown in their reference photo.
Action: energetic head nods on the beat, traded bars with clear mouth movement per line, confident posture, occasional shoulder lean toward the mic.
Camera: locked medium two-shot framing both from the waist up, slow push-in during the final two seconds.
Audio: original upbeat rap cadence, call-and-response — left performer leads, right performer answers. Write NEW original lyrics; never quote commercial songs.
Duration: 8-12 seconds, ending on a clean two-shot hold.
Template focus: AI Rap Video Duo Prompt (Two Photos).

Beat & camera breakdown

1. 0-2s — Establish. Locked two-shot. Both faces lit, the hanging mic reads clearly between them. No movement except a shared head nod on the first beat.
2. 2-5s — Left leads. Line 1 and part of line 2 land on the left performer: mouth movement, forward lean, right performer nods along silently.
3. 5-9s — Right answers. The trade reverses. Right performer takes the answering lines with the same energy; left drops back to nodding.
4. 9-12s — Push-in hold. Slow push toward the mic, both performers land a final shared pose. The hold is the loop point.

Negative prompt

identity swap, faces changing sides, merged faces, third person appearing, extra limbs, deformed hands, blurry faces, sunglasses, masks, group photo input, on-screen lyrics or karaoke text, watermarks, brand logos, celebrity lookalike, NSFW, camera shake, scene changes

Photo prep for this template

• One face per photo — no pairs, no group crops split into two uploads.
• Both photos shot from a similar angle and distance so the two performers read at the same scale.
• Distinct outfits or colors; two near-identical looks invite side swaps.
• The full photo prep checklist covers crops, glasses, and lighting.

Render steps (any model)

1. Upload photo 1 and photo 2 into your model's two reference slots, in that order — the prompt's left/right assignment follows slot order.
2. Paste the main prompt. If your tool has a separate negative field, put the negative block there; otherwise append it after "Avoid:" in the same prompt.
3. Set duration 8-12s and aspect 9:16 (or 16:9 for YouTube), then render.
4. Review the full clip against the beat breakdown. If sides drift, apply the fixes in the face troubleshooting guide — usually one wording change, not a rewrite.

Common failure modes

• Faces swap sides mid-clip → restate "fixed sides for the whole clip" and make the outfits more distinct. Full fix list in the troubleshooting guide.
• One face goes soft → that performer's photo is lower quality or smaller in frame; re-crop closer and brighter.
• Both mouths move on every line → the call-and-response roles are too vague; name which performer owns which lines explicitly in the Action block.

FAQ

Can I use three or more performers?

Two is the reliable ceiling for most image-to-video models — each extra face multiplies identity drift. For groups, render two-person pairs and cut them together.

Which performer should lead?

Whoever has the sharper photo. The lead gets more close framing and more mouth movement, so their reference does more work.