A still photo goes in. New frames come out.

What the model has to work with

The only evidence is what the pixels show: the face, the light, the pose, and whatever the frame happened to include. A video model has learned from a very large amount of footage what people tend to look like when they move. It treats your photo as the opening frame and produces frames that would plausibly follow.

A short description of the movement, such as “a slow glance toward the camera and a soft smile”, steers the model. It does not control every detail. That is why two attempts from the same photograph can look different, and why an animated portrait should never be passed off as a recording.

From your file to a finished clip

This is what happens to a Hello Once More order, step by step. Other services will differ, especially at the review stage.

1. You choose a photograph. Your device checks that it is a JPG, PNG, or WebP under 10 MB and at least 512 pixels on each side. Nothing has been uploaded yet.
2. When you continue from the review step, the page prepares a JPG copy no larger than 2,048 pixels on its longest side, uploads it, and creates your private order.
3. After payment, a person looks at the photograph and chooses a movement that suits it, such as eye contact and a smile for a close headshot.
4. The prepared image is sent to an AI video provider, which generates a silent clip of the length you paid for.
5. A person watches the entire clip for identity changes, eyes, mouth, hands, distracting gestures, and duration. If the first result is unsuitable, the included retry is used.
6. The reviewed MP4 appears on your private order page, ready to download.

Why results differ from one photo to the next

• Detail: a small or blurred face gives the model less to follow, so it guesses more.
• Pose: a face turned roughly toward the camera is easier to move convincingly than a profile.
• Hidden information: anything outside the frame or covered up, such as hands behind a back or the far side of a face, has to be invented if the movement reveals it.
• Size of the movement: a slight change of expression leaves less room for mistakes than a wave or a large turn of the head.

What animation does not do

If a head turns, the model may draw a side of the face that was never photographed, but that view is invented. It cannot recover or confirm what the person really looked like from that angle. Some tools also offer repair, color, or sound; a Hello Once More film does not include restoration or colorization, and it is silent, with no generated voice or speech. The result is an interpretation that sits beside the photograph, not a replacement for it.

If you want to know why a face sometimes drifts away from the original, read why AI-animated faces can change.