Recording a demonstration and its explanation at the same moment feels efficient. Splitting them into two passes almost always produces clearer instruction, for reasons that are mechanical rather than stylistic.
Doing and explaining compete for attention
Performing a task occupies a substantial part of working memory, particularly when the task involves precision or a sequence that must be got right on camera.
Explaining that task well requires the same resource. Doing both at once means one of them degrades, and it is almost always the explanation that suffers first.
The audible result is familiar: sentences that trail off, steps described after they have already happened, and long silences where the presenter is concentrating rather than teaching.
Separate passes let each be optimised
A demonstration recorded silently can be performed at whatever pace produces the cleanest footage, with retries costing nothing more than a second attempt.
A voice track recorded afterwards can be written, or at least planned, and delivered in a quiet room with the mouth close to the microphone rather than pointed at a screen.
Each pass therefore reaches a standard neither could reach together, and the combination is usually shorter than a single simultaneous take of the same material.
The edit gains freedom it did not have
Locked-together audio and video must be cut as one. Removing a fumbled action removes the sentence over it, which is why single-pass tutorials tend to keep material they should lose.
Independent tracks allow the visual to be sped through repetitive stretches while the narration continues at a natural pace, which is the single largest source of dead time in instructional video.
Corrections become cheap as well. A misstated step can be replaced with a few seconds of new audio rather than a complete re-record of the segment.
Ordering improves when writing comes second
Watching silent footage back before narrating it reveals what the demonstration actually shows, which is frequently not what the creator remembers doing.
Narration written against that footage describes what is on screen rather than what was intended, and the mismatch that confuses viewers disappears.
What the separate approach costs
Some spontaneity is lost. Reactions to genuinely unexpected behaviour are valuable in instruction, and a scripted voice track will not contain them unless they are deliberately preserved.
Synchronisation also takes effort, particularly where narration must land on a specific action. Marking those moments during the silent pass makes the assembly considerably faster.
For long or complex procedures the extra pass is worth it. For a short, simple demonstration the overhead can exceed the improvement, and a single careful take remains the better choice.