AI Filmmaking Short

How to Create a Dolly Zoom (Vertigo Effect) in AI Video

Master the classic Vertigo effect in AI video with a simple visual test that keeps your subject stable while background perspective distorts.

19 seconds9:16 Vertical

About This Video

What Is a Dolly Zoom?

A dolly zoom requires two things to happen at the same time: the camera moves toward or away from the subject, and the focal length changes in the opposite direction to compensate. The result is that the subject stays the same size in frame while the background behind them appears to warp — compressing if the camera moves back while zooming in, or expanding if the camera moves forward while zooming out.

The effect creates a visual tension because the viewer expects both camera movement and subject size to change together. When the subject stays fixed and the background distorts, the depth of the scene appears to stretch or collapse. This is the effect’s distinctive quality.

The One-Test Fix

The test for a correct dolly zoom is watching the subject. If the subject stays the same size throughout the shot while the background changes, the effect worked. If the subject grows or shrinks in frame, the camera and focal length are not compensating correctly.

This test applies both to evaluating prompts and to evaluating outputs. It gives a single clear criterion: subject size unchanged, background warped.

What the Prompt Structure Looks Like

Camera [moves toward / pulls back from] subject continuously.
Focal length [decreases / increases] to compensate.
Subject remains the same size in frame throughout.
Background perspective warps [toward / away from] the camera.

The compensation instruction is critical. In tests where subject-lock was not stated explicitly, outputs frequently showed the subject changing size — which means one of the two movements was applied without the other.

Observed Failure Patterns

The most common failure in Prompt to Shot tests is the single-movement output: only the camera moves, or only the focal length changes. Both produce a result that partially resembles the effect but fails the subject-lock test.

A second failure mode is timing mismatch: the camera starts moving before the focal change begins, or the two movements stop at different moments. The compensation must run simultaneously for the duration of the shot.

Testing the Result

Place a fixed reference point on the subject — a face, a shoulder, a prominent object — and track it through the output. If that reference point stays the same size from first frame to last, while the background changes around it, the shot worked. If the reference point changes size, identify which movement is out of calibration and adjust the prompt constraint.

Part of the AI Cinematic Effects playlist.

Frequently Asked Questions

What defines a dolly zoom visually?

The subject stays the same apparent size in frame throughout the shot. The background behind them warps — compressing or expanding — because the camera moves in one direction while the focal length compensates in the opposite direction. The subject lock is what makes the effect read correctly.

What is the one-test fix referenced in the title?

Watch the subject throughout the shot. If they grow or shrink in frame while the background warps, the camera and focal length are not compensating correctly. The test is simple: subject size should remain constant.

Why do AI models commonly fail to produce a dolly zoom?

In Prompt to Shot tests, models frequently produce one of the two required movements — zoom or camera movement — but not both simultaneously with the right compensation. A zoom without movement is just a zoom. Movement without zoom compensation is just a push or pull.

How should I describe the two required movements in a prompt?

Describe camera direction and compensating focal change separately. For example: 'camera pulls back while focal length increases to maintain subject size.' State the compensation explicitly rather than relying on the model to infer it.

Guides

Free & Premium Guides

Continue with the Free Guide for a practical reference, or go deeper with the Premium Guide for model-aware prompting and reference outputs.

Free Guide

Free Guide

  • FormatPDF
  • Length~28 pages
GET THE FREE GUIDE

Opens in a new tab. Your download access will be sent by Lemon Squeezy; close the checkout tab to return here.

Premium Guide

Premium Guide

  • FormatPDF, MP4
  • Length~58 pages
  • Price$14.99 USD
  • 25 effects and technique explanations
  • Model-aware prompting (Gemini, Google Flow, Kling AI)
  • 75 Reference videos (MP4)
  • Language editions: EN, FR, ES, DE, JA, PT-BR
GET THE PREMIUM GUIDE — $14.99 USD

Series

AI Cinematic Effects

Practical Shorts on cinematic effects for AI video — covering motion, timing, perspective, and camera behavior effect by effect.