Turn a still photo into a moving shot

What this is

The technique is motion transfer. Every video has camera movement baked into it — a push-in, a slow orbit, a rise. That movement can be lifted off one clip and applied to a still image you already own: a product, a logo, a storefront, a headshot.

The model reads the camera path from the source clip, works out the depth in your photo, and generates the frames in between. The result reads like you hired someone with a gimbal.

What we used

Higgsfield AI — higgsfield.ai. Inside it, the tool you want is called Genjutsu. That's the part nobody tells you: the technique is motion transfer, but the button is called Genjutsu.

It takes exactly two things — one source video whose camera movement you want, and one still image to apply it to. That's the whole input. You can render up to 1080p.

You don't need us to walk you through the buttons. Knowing what it's called and what it wants is the part that's hard to find.

What to expect

• Slow, simple moves transfer best. Gentle push-ins and orbits work. Fast, shaky or heavily-cut footage doesn't — the model needs something stable to read depth from.
• Use your highest-resolution still. This amplifies whatever is already there. A soft or compressed source comes out worse, not better.
• Text and logos are the hard case. Lettering tends to distort as the perspective shifts. Expect to try a few times if your image has type in it.
• Subtle beats dramatic. The instinct is a big move. The convincing ones are small.
• Budget a few attempts. Getting one you'd actually post on the first try is luck, not skill. That's normal.