What Makes a Good Hotel Lobby AI Photo?
Look at the face first. Can you recognize the person without zooming far into the image? Then check the framing: the photo should show a single, clearly identifiable subject, preferably with the upper body or full body visible. That matches the photo guidance in the Hotel Lobby AI generator.
If you have several photos, choose clarity over a dramatic pose or an elaborate background. A sharp everyday portrait is a more useful reference than a distant concert shot with a tiny face. These are input-selection recommendations, not a guarantee of facial accuracy in the generated video.
Use One Person Per Photo
Prepare one file for each performer. Do not upload the same group photo into both slots and expect the generator to select a different person from each copy. The upload controls identify Person 1 and Person 2; they do not let you mark a face inside a crowd.
If your only suitable photo shows both people, make two separate crops before uploading. Each crop should contain just its intended subject, with the face and as much useful body detail as possible. If cropping leaves a tiny or blurry face, find a different photo instead. Avoid combining the two crops into a collage: the workflow expects two image files.
Front-Facing vs Side Profile Photos
Choose a front-facing image when you have one. Both eyes, the nose and the outline of the face should be visible. A slight turn is a reasonable alternative when the features remain clear; a full side profile hides much of the face.
If the result does not resemble the person and the source image is strongly turned away, replace it with a more direct portrait before changing other settings. Do not try to fix an extreme angle by rotating the file: rotating changes orientation, not the camera view of the face.
Headshots vs Full-Body Photos
A full-body photo can show clothing and proportions, but it is only useful if the face is still readable. A distant full-body shot may contain less usable facial detail than a sharp upper-body portrait. When choosing between those two, use the upper-body portrait.
A close headshot gives a clear view of the face but little information about the body or outfit. Prefer a wider portrait if you want the clothing visible in the reference. Do not crop so tightly that you cut through the chin or forehead. Neither full-body framing nor a headshot guarantees that the generated clothing or body will match exactly.
Lighting and Image Quality
Check the face rather than the brightness of the whole image. A bright window behind someone can leave their face dark. Pick an image with light across the face, without deep shadows over the eyes or blown-out highlights that hide features.
Use the original image file when available rather than a screenshot of a small social-media thumbnail. Increasing its pixel dimensions cannot recover detail that is missing from the source. Likewise, selecting 720P video output does not repair a blurry reference photo.
You do not need to paint the background orange, replace the room or add a microphone to the photo. The performance template supplies the scene. Use JPG, PNG or WEBP, the formats accepted by the upload controls.
Photos You Should Avoid
- Sunglasses: choose a photo where the eyes are visible.
- Covered faces: avoid masks, hands or hair hiding major features.
- Multiple people: crop to one subject or choose an individual portrait.
- Very dark photos: replace them if you cannot clearly see the face.
- Blurred images: use a sharper original rather than enlarging the blur.
- Extreme angles: avoid looking far down, far up or away from the camera.
- Heavy face filters: use a less altered photo when recognizable likeness matters.
How to Keep the Two Performers from Getting Mixed Up
Check the previews before generating. Person 1 is the requested left performer; Person 2 is the requested right performer. If they are reversed in the upload area, use Swap to exchange them before submitting. Use separate, single-subject images so each slot has an unambiguous reference.
A left-right reversal and a blended face are different problems. If the result reverses the performers despite correctly ordered previews, the generated output did not follow the requested placement. If it blends their features, swapping the uploads alone will not supply clearer identity information. Check the references for hidden faces or extra people, replace a weak photo, and review the next result.
Clear inputs cannot lock the two identities in place. If both photos already meet the checklist, do not assume repeated cropping or increasing resolution will solve the output error. Each new generation uses credits, so inspect the inputs before submitting again.
Can You Use Old Photos?
Yes, you can try an old photo if it is clear enough to recognize the subject. For a printed image, make a straight, well-lit scan or photograph without glare across the face, then crop away the surrounding table or frame. Avoid a heavily compressed copy when a better scan is available.
Choose the age you actually want to reference. A childhood photo does not show what the person looks like today. If your goal is a present-day duo, use current photos of both subjects rather than expecting the generator to infer their adult appearance.
Can You Use Pet Photos?
For a pet experiment, use one animal per image with the eyes and muzzle visible. Choose a still, front-facing photo rather than a running shot with motion blur, and keep enough of the animal in the frame to show its shape. For two cats or dogs, prepare two separate files rather than a shared photo where their bodies overlap.
The template asks for a duo performance, so do not expect natural animal movement or exact anatomy from the result. This guide does not include a verified pet-generation test. Treat pet inputs as an experiment rather than a proven recipe for a particular output.
Before-and-After Examples
The examples below describe how to improve photo selection. They are illustrative input changes, not measured before-and-after video results.
Group Photo → Separate Performer Photos
Before: a party photo with several people and no clear intended subject. After: two separate crops, each showing one person with a readable face. Check: each upload preview should identify only its intended performer; use another photo if either crop is too small.
Distant Full-Body Shot → Clear Upper-Body Portrait
Before: the outfit is visible, but the face is tiny. After: a sharper upper-body photo with recognizable features. Check: confirm that you can see the face and clothing before submitting, rather than relying on higher video resolution.
Dark Side Profile → Evenly Lit Front View
Before: one side of the face is hidden and the visible side is in shadow. After: a front-facing photo taken in more even light. Check: both eyes and the main facial features should be visible. Review the generated likeness separately; a clearer input is not proof of a better result.
Once you have chosen both files, follow the three-step Hotel Lobby AI video tutorial for uploading, selecting settings and downloading the result.