A still photo dropped into a video has about two seconds before it stops being interesting. The Ken Burns effect is how documentary editors bought themselves eight.

What it is

Named after the documentary filmmaker whose films are largely made of photographs, it is a slow pan combined with a slow zoom across a still image. The frame moves, so the image reads as a shot rather than a slide.

The important part is that it is two moves at once. A zoom alone reads as mechanical. A pan alone reads as scanning. Together they read as a camera looking at something.

A screen recording with the camera in a corner bubble, both captured at once.

The two rules

Move toward the subject, never away from it into nothing.

If the photo is of a person standing on the left, the move should end closer to that person, or travel across the frame to arrive at them. A move that starts on the person and drifts off into empty sky is dramatically backwards: you have taken the viewer's attention and then walked away with it.

The exception, and it is a real one, is a deliberate reveal: start tight on a detail, pull out to show the context. That works because the destination is more information, not less.

Keep it under about eight percent.

Total movement, pan and zoom combined, wants to stay small. The reference is the previous rule from adding a camera move to a static video: if the viewer can see the move happening, it has become an effect rather than a shot.

Photo slideshows where every image sweeps dramatically across the frame are the failure mode. They read as a screensaver.

Duration

Four to six seconds per photo is the working range.

Under three seconds there is not enough time for the move to register, so you get a photo that twitches. Over about eight seconds you are asking a still image to hold attention on its own, and unless the photo is genuinely remarkable it will not.

If you have a lot of photos and a short video, that arithmetic decides how many make the cut. Six photos in a thirty second video is already tight.

Resolution, which is where most attempts go wrong

Zooming into an image crops it. If you push in fifteen percent on an image that is exactly your output size, the last frame is being upscaled by fifteen percent and it will look soft.

The rule: the image should be larger than your output. For 1080p output, an image of at least 2000 pixels on the long side gives you room. For 4K, aim for 4000 or more.

A phone photo is typically 4000 pixels wide or larger and has plenty of headroom. A screenshot, a logo or an image saved from a web page usually does not, and those are exactly the images that end up looking soft when they move.

Direction, per photo

Vary it. Four photos that all push in from the same corner read as a template.

A reasonable pattern for a sequence: in, across, out, in. The pull-out belongs on the last one, because ending wide is a natural close.

Also match the direction to the composition. A photo with a strong horizontal line wants a horizontal drift. A portrait wants a vertical or a straight push. Fighting the composition is what makes a move feel arbitrary.

Doing it in Zella

Photos go on the timeline like any other clip, and the move is a scale and position change across the clip's length.

On a Mac, Cinematic Move applies the push and drift in one tap, which is the Ken Burns shape by default, and you can adjust the amounts afterwards. On iPhone, set the start and end scale and position on the clip directly.

If your photos are the point of the video rather than illustrations inside it, how to make a highlight reel covers the sequencing side.

When to skip it

Charts, screenshots and anything with text. Moving text is harder to read, and a chart that drifts while someone explains it is actively worse than a static one. Hold these still.

Photos on screen for under two seconds. No time for the move to do anything; it just adds instability.

When the photo is a punchline. A reaction image or a visual joke should hit as a cut and hold. Movement softens the impact.

The Final Cut Pro documentation covers the same operation as a transform with keyframes if you want to build the move manually rather than start from a preset.

Common questions

How long should each photo be on screen? Four to six seconds. Under three there is not enough time for the move to register, and over eight you are asking a still image to hold attention on its own.

What resolution do my photos need? Larger than your output, because zooming crops. For 1080p aim for at least 2000 pixels on the long side; for 4K, 4000 or more.

Which direction should the move go? Toward the subject or across to it, never away into empty space. The exception is a deliberate reveal that pulls out to show context.