Skip to main content
Hotel Lobby AI
Generate
← Back to Blog

How to Choose Photos for a Hotel Lobby AI Duet

Prepare clear character references, assign the two opening roles, and check faces and clothing before generating your Hotel Lobby AI video.

Your photos tell the preset who should appear. Make each subject easy to distinguish before using a video take.

Use one subject per reference

Choose a separate image for each performer. If your source is a group photo, make individual crops in an image editor before uploading. Keep the other person out of each reference. Prepare these crops outside the generator.

Use the template preview to check the opening positions before assigning your photos.

Balance facial detail and clothing

Prefer a clearly visible face with even lighting and little obstruction. Do not sacrifice facial detail just to show a full outfit. A closer portrait gives less clothing information, while a wider crop can communicate an outfit. Full-body framing is not required.

Assign the starting roles

The first slot represents the performer who starts on the viewer's left. The second represents the performer who starts on the right. These are role assignments, not permanent screen positions. When a performer moves across the scene, their identity should follow the original role. Change the photo order before generating if you prefer the opposite casting.

The character-swap guide explains how the two starting roles carry through the performance.

Check both previews before submitting

Before using a take, check the following:

  • Each preview contains the person intended for that role.
  • The face has readable features and is not covered.
  • The crop includes the outfit details you want to reference.
  • The left and right uploads match the opening positions you chose.
  • You have one paid take for a 15-second, 480P video.

Keep the two original images to compare with the finished clip.

Review identity throughout the result

Compare the opening, a turn or gesture, and the ending with your references. A reversed pair from the opening differs from a face that changes during movement. On another attempt, change one reference at a time to understand the difference. Clear inputs do not guarantee an exact likeness.

Open the generator when both photos are ready. For a completed clip with problems, use the troubleshooting guide.