A speaker explains replacing a filter while the video shows arbitrary workshop images. The setting broadly fits the subject but may not help explain the procedure. B-roll becomes stronger when every shot has a specific task: identify an object, reveal an action, compare a condition or establish a location. That task also determines when the cut occurs.
Connect statements with visible evidence
Build the speech edit without extra pictures first. Mark references to objects, actions and results, then find matching footage. When a damaged part is discussed, a readable detail is more useful than a general tracking shot through the room.
The picture need not change exactly on the corresponding word. Sometimes viewers need a moment to recognise the object before hearing its explanation. Elsewhere, the statement can come first and the image confirm it afterwards. Compare both versions at normal speed instead of imposing one fixed cutting rhythm on every subject.
Build small visual sequences
A wide shot can establish the workplace, a closer view show the hands and a detail reveal the result. The sequence must remain spatially and temporally understandable. If the hand holds a different tool in the next image, viewers may infer an incorrect order. Check objects and states between shots.
Hold an important detail long enough to be understood. A complicated connection needs more viewing time than a familiar building. Duration follows information needs here, not only the music's beat.
Handle concealed edits honestly
B-roll can visually bridge an interview cut. It cannot replace missing logical context. Listen to speech alone and verify that conditions and references remain. A seamless surface does not make a distorted statement accurate.
Do not present footage of a different condition as apparent evidence of the described moment. When only general material is available, its illustrative role should remain understandable. Finally watch without sound: is the subject clear, and do the pictures complement each other? If several shots repeat the same information, a shorter selection may communicate better.