How to make it
Hotel Lobby AI Photos: How to Shoot Reference Images That Keep Faces Stable
The practical photo setup for Hotel Lobby AI videos — lighting, framing, left/right roles, and the mistakes that make two faces merge.
Updated 2026-10-05
Hotel Lobby AI videos are image-to-video jobs. The model does not invent two stable people from a prompt alone. It copies identity from the stills you upload, then performs the orange-stage choreography.
What a usable photo looks like
- One person per image.
- Face and upper body visible.
- Even light, no heavy filters, no sunglasses.
- Neutral or slightly open mouth. Extreme expressions transfer badly.
- The first upload is LEFT. The second upload is RIGHT.
Shot checklist
- Stand a few feet from a plain wall.
- Crop from mid-torso up. Do not cut the forehead or chin.
- Match the vibe of the pair — both streetwear, both formal, or both casual — without making the two people look interchangeable.
- Export a JPEG or PNG under 20 MB.
Common failures
- Selfies with a crowded background.
- Group photos used as a "single" reference.
- Two near-identical twins in the same hoodie.
- Tiny faces in a wide shot.
If the first 5-second trial looks wrong, change one photo at a time. Do not rewrite the whole setup. More failure modes are listed in troubleshooting.
Frequently asked questions
How many photos do I need for a Hotel Lobby AI video?
Two. One clear still of the left performer and one of the right performer. The generator maps those identities onto the orange-stage template.
Why do the two faces blend together?
Similar outfits, similar hair, low-resolution crops, or photos that already contain two people. Use two distinct portraits and label left versus right.
Can I use screenshots from Instagram?
Only if you own the image or have permission. Low-quality screenshots also make identity drift more likely.
Keep exploring
Ready to make your own?
Turn two reference photos into a Hotel Lobby AI video.