Diagnosing and Fixing Rolling Shutter Artifacts

Overview

Rolling shutter distortion appears in several distinct visual forms (wobble during fast pans, skewed verticals during fast motion, partial-frame flash exposure) depending on what's causing it, and the right fix depends on correctly identifying which specific form you're seeing.

What You Need

• Whatever gear produced the problem. This is a diagnostic guide, not a gear-buying tutorial

Steps

Pro Tips

• Use a faster shutter speed or camera stabilization to reduce wobble and skew artifacts during fast movement. This addresses the cause rather than relying on post-production correction alone.
• If shooting with flash or LED lighting, test your shutter speed against that specific light source before a real shoot, banding and partial exposure issues are much easier to prevent than to fix afterward.
• Check whether your camera offers a global shutter mode or a different sensor readout speed setting for situations where rolling shutter artifacts are a known problem for your specific shooting style.

Tested: what sets the lean, and what a shutter speed can and cannot fix

We tested steps 1 to 3 and Pro Tips 1 and 3 on 2026-09-29 with a model of a rolling shutter written in numpy. It is a model, not a camera. Each row of the frame is read out a little later than the row above it, and each row collects light for the shutter time from the moment it starts. The scene is a fence of vertical bars of different widths on a frame 960 pixels wide, and the camera pans past it at a steady speed. We measured the position of every row against the known pattern, and the frame period is 1/30 of a second (33.3 ms).

Steps 1 and 2 hold, and the lean has a simple rule. Across the height of the frame the bars leaned by pan speed times readout time: 3.2, 6.4 and 12.8 pixels at 4, 8 and 16 pixels per frame with a readout taking 0.8 of a frame, and 1.3, 2.4 and 4.8 pixels with a readout taking 0.3 of a frame. Every measured value was within about 15 percent of that product. Halving the pan speed halves the lean, and a sensor that reads out in a third of the time cuts it to a third.

Pro Tip 1 does not hold for a faster shutter speed, in this model. We rendered the same pan at shutter times of half, a quarter and a tenth of a frame. The lean stayed at 6.4, 6.4 and 6.4 pixels. What changed was blur: edge sharpness rose from 47 percent of a still frame at the slowest shutter to 76 percent at the fastest. A faster shutter sharpens each row and leaves the timing between rows alone, because that timing belongs to the sensor's readout. In this model only the pan speed and the readout time move the lean, which supports the page's advice to stabilize and to look for a faster readout setting in Pro Tip 3.

Step 3 needs one more condition than the page gives. We placed a 4 ms flash at every 0.1 ms position in a frame period and counted the positions where one frame's every row was exposing for all of it. With the readout at 0.8 of a frame, a shutter of 1/30 second, the slowest available at 30 frames a second, lit the whole frame evenly for only 8 percent of flash positions. The flash has to land after the last row has started exposing and before the first row stops, and that window is the shutter time minus the readout time, with the flash's own length still to fit inside it. With the slow readout, shutters of 1/60 second or faster never lit a whole frame. Cutting the readout to 0.3 of a frame raised the 1/30 second case to 58 percent, which is the case for Pro Tip 3's readout mode.

Validated in a model. Measured 2026-09-29 with numpy. This is a simulation of how a rolling shutter reads out, not footage from a camera, so real sensors, lenses and flash tubes will differ. We did not test rolling shutter correction in any editor, and FFmpeg has none. Its deshake filter stabilizes whole frames and drifted on this synthetic footage, so we did not use it. The pan speeds are for a 960 pixel wide frame. Flash positions are sampled every 0.1 ms for a 4 ms flash at 30 frames a second.