The five-second photo checklist

Check each image before uploading. The model can only use the visual information present in the reference, so a recognizable photo on a small phone screen may still be a weak input when the face itself contains very few clear pixels.

  • One adult per photo, with permission to use it.
  • Face and hair fully visible, without sunglasses or a hand covering them.
  • Front-facing or gently angled rather than a strong side profile.
  • Even light, natural skin detail and no obvious motion blur.
  • Chest-up or waist-up framing with some space around the head.

Choose a clear face angle and expression

A front-facing portrait or a gentle three-quarter angle gives the generator more visible information about both sides of the face. A strong profile hides one eye and part of the facial structure, which can make identity consistency harder during movement.

Use a relaxed, natural expression when possible. An extreme expression can become part of the reference and may not transfer cleanly into a performance with different mouth and head movement.

Keep natural detail and even lighting

Choose a photo where the face is neither lost in shadow nor washed out by a bright light. Soft daylight or balanced indoor light is usually easier to read than colored club lighting, a bright window behind the subject, or harsh shadow across half the face.

Avoid beauty filters that reshape the eyes, jaw or skin and avoid aggressive portrait blur that removes the edges of the hair. The generated result may already vary; a heavily altered reference adds another source of uncertainty.

Crop for the person, not just the scene

The face should occupy a useful part of the image. A full landscape photo can look sharp while the person inside it is too small to serve as a strong identity reference. Crop closer without cutting off the chin, forehead or hair.

Do not repeatedly save, message and screenshot the same photo. Each extra copy can remove detail. Use the closest available version to the original image, within the upload formats and size limits shown by the Studio.

  • Good starting point: chest-up or waist-up portrait.
  • Keep the full head, hairline, chin and both eyes in frame.
  • Do not enlarge a tiny face until it becomes blocky or smeared.
  • Use the original file instead of a screenshot when available.

How to use a photo of both people

The duet workflow expects a separate reference for each person. If your only suitable source is one photo together, make two copies and crop one person from each copy. Each crop should contain one complete face and enough hair and shoulders to remain readable.

A shared photo works only when both original faces are large and sharp enough. If either person is far from the camera, covered, blurred or turned away, choose a different individual portrait for that role.

The two portraits do not need to match

The photos can come from different places, devices and dates. Matching backgrounds are unnecessary because the prepared performance supplies the scene. What matters more is that each person's identity is easy to see.

Similar image quality and neutral lighting can reduce avoidable differences, but do not edit the two people into an artificial matching style. Preserve natural facial information instead.

If the result does not look like you

First review the reference rather than assuming a second submission will fix the result. Check whether the face was small, filtered, partly covered, strongly angled or blurred. Replace only the weak reference and verify that the two roles have not been swapped.

A clearer photo improves the information available to the model, but it does not guarantee an exact face, expression or frame-by-frame identity. A new generation is a new creative output and may use additional credits. Follow the published review and refund process for a completed result with a quality concern.

Use only photos you have permission to process

A technically good photo is not automatically an appropriate photo. Use your own image or obtain permission from every person whose likeness you upload. Do not treat a public profile, repost or search result as consent.

Review the privacy policy before uploading. It explains site storage, deletion, external AI processing and why the site's retention period is separate from a provider's terms.

Common questions

What kind of photo works best for Raindance AI?

Use a clear, recent chest-up or waist-up portrait of one consenting adult. Keep the full face and hair visible in even light, with a front-facing or gently angled pose.

Can I use a selfie?

Yes, if it is sharp, naturally lit and does not distort or cover the face. Avoid extreme wide-angle distortion, strong filters, sunglasses and hands over the face.

Can I crop two people from one photo?

Yes. Make a separate crop for each person, keeping the complete face and hair. Choose different photos if either face is too small, blurred, covered or turned away.

Do the two photos need the same background or lighting?

No. They can come from different places and cameras. Clear identity detail matters more than matching backgrounds, although balanced lighting and comparable image quality can help.

Will a better photo guarantee an exact likeness?

No. A clearer reference gives the model better information, but AI can still change facial details, expressions, roles, framing and motion.

Why does my Raindance AI video not look like me?

Check for a tiny face, blur, heavy filters, a strong profile, facial obstruction or swapped role assignment. Replace a weak reference before considering another generation, while remembering that exact likeness is not guaranteed.

Sources and review

Reviewed on . Sources cover the music and platform context. Check the generator for current availability, output settings, and credit costs.