The best AI duet photos are two separate shots, one person or pet in each, facing the camera at eye level in even daylight, with nothing covering the face.
- Frame: head and shoulders in the shot.
- What fails: group shots, sunglasses, side profiles, and dark or blurry photos.
- Files: JPG, PNG, WebP or iPhone HEIC, at least 300 pixels on each side.
The AI Duet then puts you both side by side in one 15-second shot, rapping in turn, for $9.90 per video.
Why the photos matter so much
The AI Duet builds both rappers from your two photos. Each face, hair or fur, skin tone and outfit comes from its own photo. The video model has to animate two mouths, in turn, to two different voices. A sharp, front-facing face gives it a clear mouth to sync. A turned or covered face leaves it guessing, and that's where likeness and lip-sync slip first.
What works and what fails
| Photo | Result | Why |
|---|---|---|
| Selfie or head-and-shoulders, facing camera | Works best | Clear eyes and mouth, easy to keep the likeness |
| Waist-up, facing camera | Works | Matches the waist-up framing of the video; keeps your outfit |
| Daylight by a window, even light | Works best | Skin tone and features stay true |
| Pet at eye level, face filling the frame | Works | Breed, colors and markings carry over |
| Slight smile, mouth visible | Works | Gives the model a natural mouth to animate |
| Two people in one photo | Fails | One photo must hold one subject; the AI may mix up faces |
| Sunglasses, mask, hand over the face | Fails | Covered features get invented |
| Side profile or looking far away | Often off | The video faces the camera; a profile has half a face to work from |
| Dark, backlit or night photo | Often off | Features get lost; dark pets suffer most |
| Blurry, tiny or heavily zoomed crop | Fails | Below 300 px is refused; soft photos lose likeness |
| Heavy beauty filter or cartoon filter | Often off | The filtered face is what gets animated |
| Big hat, hood or headphones | Mixed | What you wear in the photo stays on, and can hide the hairline or ears |
| Celebrity, real rapper or a child | Not allowed | Only photos you have the right to use, of adults or pets |
Matching the two photos
The two photos don't need to match, but a pair that looks alike reads better side by side:
- Similar framing. Two head-and-shoulders shots, or two waist-up shots. A tight selfie next to a full-body shot can look uneven.
- Similar light. Two daylight photos sit together better than one bright and one dim.
- Clothes you like. Each person keeps the outfit from their photo, so this is your costume choice.
- Order matters. Your photo goes first: you stand on the left and rap verse 1. Your partner's photo stands on the right and raps verse 2.
Pairs that work well
- Couple, two selfies from the same evening. Same light, same mood. For the song, borrow lines from our rap duet song for couples.
- Best friends, two waist-up shots. Outfits show, so it feels like a real music video.
- You and your dog: your selfie plus a photo of your dog taken at its eye level. See how an AI duet with your pet works.
- Mom and you: two front-facing portraits, both smiling.
Before you upload
- Crop each photo so one face is clearly the subject.
- Zoom in and check that the face is sharp.
- Make sure both people agreed to be in the duet.
- Write the lyrics and hear the song first. Lyrics are free (up to 30 AI drafts a day), and so is your first song each day. You only add photos and pay once the song is right.
If a photo is refused, it's usually too small (under 300 px) or not a JPG, PNG, WebP or HEIC file. Swap it before you pay.
FAQ
Can I use one photo with both of us in it?
No. Use two photos with one subject each. You can crop a group photo into two if both faces are sharp and large enough.
Does the photo need a plain background?
No. The background is replaced by the scene you pick, cozy studio or orange lobby.
What if the face doesn't look right?
Try a sharper, front-facing photo in even light. If a video fails, its credits come back automatically, so you can make it again at no extra cost.
Can I use a photo of a friend without asking?
No. Only use photos of people who agreed. No celebrities, no real rappers, no minors.
How long does it take?
Usually about 5 minutes for the 15-second video, and we email you the link too. For every step, see how to make an AI duet video.