Top-down shots with AI: straight down, no distortion

A top-down shot, also called a bird's-eye view or overhead shot, is a frame where the camera sits directly above your subject and looks straight down, at roughly a 90-degree angle. It's one of the hardest perspectives to make with AI, because image models almost always fall back on a slight tilted angle instead of a truly perpendicular view. Here's why that happens and how to get it anyway.

What exactly is a top-down shot?

A top-down shot looks straight down on your subject, while a high-angle shot only looks down at an angle. That difference changes everything. With a high angle you keep depth and space, because you still see the sides of objects. With a true top-down frame that depth disappears and you only see the tops, laid out flat.

Film guides like StudioBinder describe the overhead shot as a camera placed directly above the scene, at roughly 90 degrees. Move the camera much farther away and you're closer to an aerial or God's-eye view. Director Wes Anderson even turned the straight-down insert shot into a signature: flat, symmetrical images of objects on a table, shot from directly above.

For content this perspective is gold for flat lays: products, dishes, accessories or a laid-out outfit, neatly arranged and framed straight from above.

Why AI would rather tilt the camera

AI image models tend to turn a top-down request into a tilted downward angle, because their training data is full of eye-level photos and holds far fewer truly perpendicular shots. The model simply picks the perspective it has seen most often.

Ask for "from above" and the model happily reads it as "a bit higher than eye level", not a strict 90 degrees. You'll spot the result right away: you can still see the side of a cup or the back of a chair, when you wanted a flat, geometric composition.

The fix isn't pushing harder with the same word, but describing more precisely what you mean.

How to force a truly perpendicular view

Name the angle explicitly and leave no room for interpretation. Combine several terms that describe the same perpendicular perspective, so the model doesn't fall back on its average.

• Use terms like top-down, directly overhead, shot from directly above, 90-degree overhead and flat lay.
• Describe the plane: objects flat on the table, camera parallel to the surface, no visible sides.
• Say what's there and how it's arranged, because a top-down image lives on the composition across the flat plane.
• State explicitly what you don't want: no tilt, no perspective, not a high angle.

Not sure how to phrase this compactly? Let the prompt generator turn your loose idea into a clean prompt with angle, composition and light built in. Then generate the image in the photo generator.

Avoiding distortion: keep it flat and perpendicular

A top-down image only looks right when the camera is truly parallel to the surface, because any tilt causes keystoning: lines that converge and objects that skew away. Circles then turn into ovals and rectangles into trapezoids.

Flat-lay photography guides treat an angle between 85 and 90 degrees as the range that reads as distortion-free to the viewer. Below that, everyone can tell the frame is off. For AI this means: the clearer you describe "perfectly flat, straight down, no tilt", the smaller the chance of a skewed edge.

If you still see distortion or an object that tips away oddly, fix it selectively with inpainting instead of regenerating the whole image.

Top-down in motion: the flat lay as video

A moving top-down shot works the same way, but asks for extra calm in the motion. Think of hands entering the frame from the side, a slow push-in straight down, or ingredients appearing on the surface one by one.

Keep the camera perpendicular and let only the subject or a hand move, not the whole frame tilt. Describe the motion small and concrete in the video generator, and generate a still as a start frame first so your angle is locked before you bring it to life.

Frequently asked questions

What's the difference between a top-down shot and a high-angle shot?

A top-down shot looks straight down at roughly 90 degrees and shows only the top of your subject. A high-angle shot looks down at an angle and keeps depth and sides visible.

Why does AI turn my top-down image into a tilted angle?

Because the training data is mostly eye-level photos and far fewer perpendicular shots. The model falls back on the perspective it saw most often, unless you describe the 90-degree angle explicitly and with several terms.

Which prompt words work for a straight-from-above image?

Combine top-down, directly overhead, 90-degree overhead and flat lay, describe the flat plane and composition, and rule out what you don't want with no tilt, no perspective.

Can I make a top-down shot as video too?

Yes. Keep the camera perpendicular and let only the subject or a hand move. Generate a top-down still as a start frame first and then bring it to life, so the angle stays fixed.

The top-down perspective stands out because it's rare and easy to read, exactly what makes a flat lay or insert shot so strong. With an explicit prompt and a flat, perpendicular frame, you can get it with AI too. Create an account and try your first top-down image.