Good portraits are a better starting point.

One clear adult face in each file

Choose two separate portraits, each showing one adult. A head-and-shoulders image with the eyes visible makes it easier to see whether you selected a useful reference. Avoid a group image where several people might compete for attention.

Do not crop a tiny face from a distant crowd photograph and assume enlarging the file will restore detail. Start from an original image with enough real facial detail. A minimum of 240 × 240 pixels is required, but the minimum is an acceptance limit rather than a quality recommendation.

Keep the features readable

Soft, reasonably even light is a practical starting point. Check that neither side of the face disappears into shadow and that a bright window has not turned the person into a silhouette. Heavy glare, sunglasses and objects covering the mouth remove useful information.

Strong beauty filters or stylized edits can change the reference before the video system sees it. If recognition matters, prefer a recent, lightly processed image. You do not need to recreate the final scene in the portrait: the selected scene supplies that direction.

Check the file, not just its thumbnail

• Accepted formats: JPG, PNG and WebP.
• Minimum dimensions: 240 × 240 pixels.
• Maximum file size: 10 MB per image.
• Maximum image area: 40 megapixels per image.

If your phone saves a photo in another format, export a compatible copy before uploading. Renaming a file extension is not a format conversion. Open the exported file once to make sure the image still looks correct and is oriented as intended.

Agree on how the portraits will be used

Ask both people before making an AI performance with their faces. Being sent a photo for one purpose does not automatically mean its owner wants it turned into a music video or posted publicly. Agree on whether the result is just for a private chat or may be shared more widely.

Photo selection initially previews the files in your browser. Confirming generation uploads them for processing. The privacy information explains storage and provider processing; private delivered videos remain available for 30 days.

A final check before confirming

1. Is this the intended adult in each preview?
2. Can you clearly see the eyes, face shape and mouth?
3. Do both people agree to this use and the intended sharing?
4. Does your chosen scene fit the occasion?
5. Have you watched a real example and accepted that facial detail, hands and lip sync may vary?

When the inputs are ready, use the photo-to-rap walkthrough. After delivery, repeat the visual check on the full moving video, not only its opening frame.