Can I use a selfie for both photos?

You can use a selfie for the close-up role, but the hallway role needs a full body with visible legs or the sink does not read.

Does image quality matter more than framing?

Framing matters more. A sharp photo with the head too small fails where a slightly softer photo with the right crop succeeds.

What if I only have one good photo of each person?

That is enough. One photo per person is the input; you do not need front and side views.

Will the model keep their clothes?

It keeps the clothes from your photos as the starting point, though patterns and small details can change during the motion.

Short answer

Use one adult per photo, face to camera, no sunglasses or hat, even light. Photo 1 needs a face large enough to survive a close-up; photo 2 needs visible legs, because the sink is the joke. Both photos should be the same kind of framing. Group shots, covered faces and heavy filters reliably break the render.

Make my band video

Hop Hop Hop AI photo guide

The clip asks one face to be enormous and the other to be tiny. Only some photos survive that.

Updated 2026-11-02

Why these photos fail more than most trends

Most photo-to-video formats keep one person at roughly one size for the whole clip. This one does the opposite: it takes a face and blows it up to fill a third of the frame, then cuts to a full body a few metres away, then swaps both. A photo that looks fine on a profile can still be the wrong raw material here.

Photo 1 — the pointer (close-up first)

Face to camera, eyes open, nothing across the eyes.
Head fills a decent share of the frame — a distant full-body shot on a wide street will not survive the close-up.
Shoulders and upper body visible, so the pointing arm has somewhere to come from.
Neutral or slight smile. A huge open laugh is fine, but it will not be reproduced exactly.

Photo 2 — the bouncer (hallway first)

Full body, head to feet, with the legs clearly visible.
Standing, weight even, facing the camera.
No props crossing the body — the model has to fold this figure in half.
Same rough lighting as photo 1. Mismatched light is the most common reason one of the two looks pasted in.

Things that quietly ruin a render

Input	What happens
Sunglasses	Dark eye-wear reads as a shape and gets re-drawn onto the new face
A hat, hood or headscarf	The head silhouette is copied even when the face is not
Two people in one photo	Ambiguous which body to replace
Screenshot with UI or stickers	Text and borders bleed into the frame
Heavy beauty filter	Skin texture is invented, and the face drifts toward an average
Minors or public figures	Rejected by moderation before any render starts

Lighting, and why it is not fussy advice

The clip reconstructs one corridor with one light source. If your two photos were shot under a window and under a ceiling bulb, the model has to reconcile them, and the person it understands less well is the one that drifts. Shooting both photos in the same room, facing the same way, removes an entire class of failure for free.

How to check a photo before you spend credits

1Zoom to the face. Is it sharp at twice the size you see now? If not, the close-up will be soft.
2Check for eye-wear and head coverings. Remove them.
3Confirm the far person's feet are in frame.
4Compare the two photos side by side. Do the people look like they were lit in the same place?
5Confirm both are adults who have agreed to appear.

Consent is not paperwork here. You are putting someone's face into a video that will be posted publicly, labelled or not — they should know that before you upload, not after.