YOUR REFERENCES DO THE FIRST JOB

Choose Better Photos for an AI Duet

A sharp, unobstructed face gives the generator useful detail. A group screenshot or a heavily filtered selfie gives it ambiguity. Use this checklist before spending credits on a new take.

ACTUAL GENERATED EXAMPLE

Two Photos to an Orange-Booth AI Duet

Watch a 10-second, 720p orange-booth promotional duet made from the two displayed fictional adult references, with an original hip-hop groove and alternating sung vocals. Individual results vary.

A curated promotional example using the displayed references and custom performance direction. Exported to a 10-second 720p MP4 with its native generated soundtrack; no added music or visual retouching. The public studio offers its own generation profiles and credit quote.

See the inputs and full video ↗

Use one visible adult per upload

Give each slot a separate image with one subject. Avoid photos where a second face, poster or mirror reflection competes for attention. Both adults must permit the transformation and your intended use of the result.

Keep the eyes, mouth and jawline visible. A front-facing photo or gentle angle is easier to compare with the result than a face hidden by a hand or extreme profile. Choose even light rather than a strong silhouette or dark party shot.

  • One clearly visible adult in each image.
  • A sharp original with a readable face and natural lighting.
  • No strong beauty filter, motion blur or face-covering glasses.
  • Permission for both the photos and intended publication.

Match framing and preserve useful detail

Two similarly framed portraits make the pair easier to judge. Include enough head and upper-body detail to guide hair and clothing, while keeping the face large enough to remain readable. A high pixel count does not help if most of the image is an empty background.

Use the original image file when possible. Cropping a low-resolution screenshot or repeatedly re-saving it can remove facial detail. Better references reduce ambiguity, but cannot guarantee perfect identity or movement.

Check the actual file requirements

The studio accepts JPG, PNG and WebP. Each file must be no more than 10 MB and at least 256 × 256 pixels. Performer 1 is the left reference and Performer 2 is the right reference. Verify both slots before you confirm permission and start.

Use a short Preview to assess a new pair, then decide whether to generate a longer Standard take. Each new generation has its own credit cost; a completed test does not turn into a free upgrade.

A few things to know.

Can I upload a group photo into both slots?+

Use one clear adult per separate image. Multiple faces create ambiguous reference identity.

Will a larger file guarantee a better face?+

No. Visible facial detail, lighting and framing matter; larger dimensions cannot restore a blurred face.

Can I use a photo without the person's permission?+

No. Both people must be consenting adults and you must have permission to transform the photos.

Are sunglasses or heavy shadows useful references?+

They can hide facial details that matter for resemblance. Prefer a readable face with even lighting and enough detail for the hair and outfit.