Which Photos Work Best for Hotel Lobby AI Videos?
Short answer: use one photo per person, face toward the camera in good light, ideally head to toe and standing, with nothing covering the face and nothing in the hands. On Wan 3.0, each photo is first redrawn as a full-length studio portrait of the same person, so whatever your photo does not show, the redraw has to invent.
This checklist comes from how the generator handles photos and from what we saw in our own tests between October 2 and 4, 2026. Those tests used photos we cannot publish, so we describe them in words; a side-by-side photo comparison is not published yet.
Updated
HOTEL LOBBY AI · PHOTO GUIDE
Pick photos that keep the face
- One per photo
- Head to toe
- Eyes visible
The quick photo checklist
| Pick | Avoid | Why |
|---|---|---|
| One person per photo | Group shots, a friend cut off at the edge | An extra face leaves the model a choice to make. Crop to one person. |
| Face toward the camera, evenly lit | Profiles, dark rooms, strong light behind | The face is what has to survive the redraw and the video. |
| Head to toe, or at least to the waist | Face-only selfies | The redraw fills in the missing body and clothes with plain items that match the top. |
| Standing | Sitting or lying down | The redraw stands everyone up, but it has to guess the legs and the posture. |
| Mouth closed or a soft smile | Tongue out, a wide-open mouth | Open mouths looked odd in our tests; the redraw now asks for a closed mouth. |
| Empty hands | Phones, drinks, bouquets | The redraw drops held items and some small jewelry. |
| Eyes visible | Sunglasses, masks, a cap pulled low | Covered eyes leave the model nothing to copy. |
| The original file, as large as you have | Screenshots of chats, tiny copies | Photos under 300 pixels on a side are refused; blurry ones lose detail. |
Photos ready?
Try your photos in the booth
One person per photo, the left raps and the right hypes. The button shows the credits before you create.
Your two performers
One person each. Left raps, right hypes.
Stage
Video & sound
This video: 6 credits. Ready in about 5 to 15 minutes; failed generations return your credits. View pricing
What happens to your photo before the video
- In your browser, the photo is redrawn as a JPG (longest side 2,048 pixels), and its location data is removed.
- On Wan 3.0 in the Hotel Lobby booth, each photo is redrawn as a full-length studio portrait: same face, hair, age, skin or fur and body shape; standing, arms relaxed, mouth closed, plain gray background. Pets stand on their hind legs.
- Parts the photo does not show are filled in with plain clothes that match what is visible. Jewelry, glasses and hats are not added.
- The video model then performs the booth routine with the two portraits.
- You never see the portrait. It is deleted with your photos 7 days after the render. If a portrait fails or takes longer than 5 minutes, your original photo is used.
The redraw does not run with Your clip or on Seedance models. There, your photos go in as they are, so full-length photos matter even more.
Two photos or one photo of both?
| Mode | How it works | Good to know |
|---|---|---|
| Two photos | One person per slot; the left raps, the right hypes | The most control over who goes where. The photos do not need to be taken together. |
| One photo | Both of you side by side, faces clear; the left of the photo raps | On Wan 3.0 the photo is split into a left and a right portrait. Very wide photos are refused. |
We have not published a test of the one-photo route yet. If you have separate photos of each person, use Two photos.
What we saw in our own tests
- A dog photographed sitting stayed seated in an early test, before the redraw step existed. With the standing redraw, pets now get onto their hind legs.
- A tongue-out photo carried the tongue into the portrait until we asked the redraw for a closed mouth.
- Babies kept their faces but looked a little older once they were stood up.
- Dog paws turned hand-shaped when the routine points or gestures.
- Sizes are not kept to scale: a small dog stands as tall as an adult.
- The same two photos give a slightly different video each time.
The full log of what went wrong and what we changed is in our test results.
What better photos cannot fix
- Likeness is never guaranteed. Faces can drift, most often during fast turns.
- Moves, camera and timing come from the routine, not from your photo.
- A higher resolution makes the video sharper. We have seen nothing showing it makes a face more like the photo.
- Each new try with a better photo costs credits again; a render that fails is refunded automatically.
Photos of other people: ask first
Use photos of people who agreed to be in the video. TikTok does not allow AI content showing anyone under 18, or adults who are private people without their permission, and asks you to label realistic AI content (TikTok). The trend itself has drawn debate over AI copies of Takeoff, who died in 2022; Quavo told the AP on October 2, 2026: “I’m proud of it. I thank you for keeping my brother and his name alive.” (AP via WSLS).
FAQ
Can I use a selfie for a Hotel Lobby AI video?
Yes, if the face is clear and front-facing. The redraw will invent the body and the outfit, so a photo to the waist or head to toe keeps more of the real person.
Do the two photos need to be taken together?
No. Two separate photos, one person each, is the usual way. They can come from different days and places.
Can I use a photo with sunglasses?
It is accepted, but the eyes behind the glasses cannot be recovered. A photo with the eyes visible gives a closer face.
Why does my outfit look different in the video?
The redraw keeps what the photo shows and fills in the rest with plain clothes that match. Held items and some accessories are dropped.
Can I use a photo of a baby or a child?
The generator accepts it, and our tests kept babies’ faces. Use it only with a parent’s or guardian’s consent, and note that TikTok does not allow AI content showing anyone under 18.
Does a bigger photo help?
Up to a point. Photos are resized to 2,048 pixels on the longest side in your browser; a sharp, well-lit photo matters more than a huge one.
Sources
Read next
How to Do the Hotel Lobby Trend: 3 Ways, Step by Step (2026)
Three ways to do the Hotel Lobby AI trend (two-photo generator, AI video prompt, editor template), with steps, costs, free options and tips
Hotel Lobby AI Prompts: Our Exact Prompt and 5 to Copy
The exact Hotel Lobby AI prompts our generator sends (Wan 3.0 with the Hotel Lobby track, and Seedance with an AI rap), what each line does, and five photo-only prompts to copy for couples, friends, family, characters and pets
Hotel Lobby AI Video Results: Our Tests, Fixes and Limits
An honest test log: swapped faces, copied gloves, cropped performers and blocked songs, what we changed after each, and what still varies
Hotel Lobby Video is an independent tool, not affiliated with Quavo, Takeoff, Migos, Quality Control Music, Motown or COLORS.