Hotel Lobby AI Photo Tips: Faces, Crops & Left vs Right

For a two-person Hotel Lobby video, start with two separate, sharp photos with one adult in each. Keep the whole face visible and deliberately assign each photo to the left or right performer. Then inspect the generated still scene before animating it. These are useful checks you can control; they do not eliminate every identity or motion error.

FunClip accepts JPG, PNG and WebP images smaller than 10 MB each. If your phone saved an HEIC photo, export a JPG or PNG copy first. Renaming the file extension does not convert the image.

Why the two-photo setup matters

A public discussion about making the Hotel Lobby effect explains how easily the setup becomes complicated: the creator distinguishes the left person, right person, close-up and full-body references. The practical question is simple: how does the system know which face belongs to which performer?

In FunClip, Left performer photo and Right performer photo answer that question explicitly. The first scene is built around those assignments. If you place the photos in the opposite slots, the intended casting changes. Upload order is worth checking before you pay for an extra version.

Choose the useful photo, not necessarily the favourite photo

Your favourite picture might be a night-out selfie with a hand across the face. For this task, a plain daylight portrait can be more useful. You want to see the eyes, nose, mouth, jaw and hairline without having to guess what is hidden.

Open each photo at a comfortable viewing size. If you cannot clearly see a person's features, enlarging the file will not recreate that missing information. Look for an original camera image instead of a screenshot that has already been compressed several times. A sharp head-and-shoulders photo is a better starting point than a distant full-body shot in which the face occupies only a few pixels.

Avoid strong beauty filters, extreme coloured lighting, oversized sunglasses and hair covering most of the face. These are practical ways to reduce ambiguity, not a scoring system. FunClip does not currently give your upload a photo-quality score or promise an identity match percentage.

If the only photo you have is a group picture

Crop it into one image per intended performer before uploading. Keep a little space around the head and shoulders. Compare the crop with the original: if the person was very small in the group, the crop may be too soft to help.

Do not treat the Solo, Duo and Party pack names as a person selector. Every current Hotel Lobby clip is built for two adults, and a Party pack contains five video allowances. It does not turn a photograph of five friends into a five-person performance. If your task needs a whole group in one frame, the current template is not the right fit.

For example, imagine a photograph of four friends at dinner. You want the two adults at either end. Make one crop of the leftmost person and a separate crop of the rightmost person, leaving the other faces outside those images. If either crop is blurry, ask that friend for another portrait rather than adding a paragraph of instructions: this template does not offer a custom prompt box.

Give the two inputs a fair chance to work together

You do not need identical backgrounds, cameras or outfits. Two independently taken portraits can work. When you have a choice, use photos with similarly clear faces and reasonably even light. One very dark profile beside a bright front-facing portrait gives you a harder pair to evaluate.

The scene generator aims to preserve the two people and their outfits while composing them in the orange studio. It may infer body details that are outside the source photos. A face close-up therefore cannot guarantee an exact full-body outfit. Avoid promising someone a precise uniform, branded garment or jewellery detail based on a photo where it is not visible.

Check the still scene before spending the video step

Use a short review sequence:

1. Sides: is the intended person on each side of the microphone?
2. Faces: does each person look recognisable enough for the joke or message?
3. Framing: are both people inside the frame, with no important area clipped?
4. Details: are hands, hair and clothing acceptable at normal viewing size?

Choose 9:16 or 16:9 according to the final destination before creating that scene. A portrait composition and a landscape composition give the model different space to arrange the pair. Changing your mind after a successful generation means creating a new version, not recovering a failed step.

If one face is weak, replace that input with a clearer portrait. Keep the stronger input unchanged so you can judge whether the replacement helped. Repeating the exact same poor inputs can produce a different result, but it does not address why the starting material was difficult.

To start over after the still scene has finished, reopen the home page at its clean URL and leave Restore scene unselected. Upload your photos into the new form. The current scene cannot be edited in place. Starting a new scene uses additional allowance and may require another purchase; the successful old scene is not refunded. Only do this once the current task has finished, and check account history if its status is uncertain.

A good still image does not guarantee a perfect video

Motion introduces another check. Watch both faces throughout the 10-second result, especially when a head turns or a hand passes near a face. Listen to the audio and inspect mouth movement without assuming exact lip sync. If a completed result is not to your taste, a new version uses allowance; technical failure recovery is a separate process.

Use the full walkthrough when you are ready to create. If the account shows an unfinished task, check My clips before starting again. For billing boundaries, read the one-time pack guide.

Choose your two photos