Frame-motion analysis catches handheld instability, camera vibration and tripod slippage. Every unstable section gets a severity-ranked marker so you can decide: stabilise, cut or replace with B-roll.
Motion instability takes different forms depending on shooting conditions. EditLint AI identifies all three categories with separate severity thresholds.
High-frequency, low-amplitude vibrations caused by holding the camera freehand. The detector measures inter-frame motion vector variance — a handheld signature produces a distinctive noise pattern across multiple axes that distinguishes it from intentional camera movement or panning.
Higher-frequency shake caused by vehicle movement, machinery nearby, or DSLR mirror slap. Vibration produces a periodic oscillation pattern in the motion vectors that differs from organic handheld shake. Detected as medium or high severity depending on amplitude and frequency.
Slow, low-frequency drift caused by a pan head that wasn't fully locked, or a head that slipped slightly during recording. Unlike handheld shake, this produces a slow unidirectional drift in the motion vectors that's often invisible to the eye on playback but noticeable to viewers subliminally.
Not all shake is equal. A tiny handheld wobble on a wide shot is far less noticeable than the same motion on a tight close-up. EditLint AI's severity scoring factors in both the motion amplitude and the focal length metadata so you see an accurate picture of real viewer impact.
Severity thresholds are adjustable per project. Broadcast and documentary editors often lower the threshold; run-and-gun and vlog editors often raise it.
For every clip in the selected range, EditLint AI samples frames at regular intervals and computes dense optical flow between consecutive frames. This produces a motion vector field that captures the direction and magnitude of movement across the entire frame.
The motion vector fields are analysed for patterns characteristic of shake, vibration and drift. A rolling instability score is computed for each second of footage. Sections where the score exceeds the configured threshold for more than the minimum duration are flagged as candidates.
Each flagged section gets a colour-coded [EditLint] marker placed at its start point on the clip's timeline. High severity in red, medium in amber, low in cyan. The marker comment includes the detected shake type, average severity score and a suggested action.
A 60-second full-sequence scan catches every unstable section with severity ranking so you fix what matters first and skip what doesn't.