Scene length and inserted scenes

What it does

An ordinary scene is as long as its narration turns out to be. You can hold it longer — a beat after the voice stops — but you cannot make it shorter than the line, because a shorter scene is a sentence cut in half.

An inserted scene has no narration and declares its length outright — a sponsor, a product shot, a title card. Drop one in and the rest of the video moves down by exactly that much.

The standard behind it

• Authoring can extend a scene and never truncate its narration. The scene holds for the longer of the speech and the number you set: ask for 4.4 seconds over a 4.0-second line and the shot runs on for four tenths after the voice stops; ask for 2.0 and the scene holds at 4.0, and the control says so rather than obeying. It is the same shape as the backdrop, which dims further for taste and never lighter than legibility requires.
• A sponsor is a scene, not an overlay. It occupies the timeline, so the viewer scrubs past it with their player’s own controls like any other part of the video.
• Everything after it moves, including every caption and banner, because those are anchored to a scene and an offset rather than to a clock. A banner placed "at 0:13" and stored as 13 seconds would annotate a different moment the instant anything was inserted in front of it.
• A scene with neither narration nor a declared length is refused. It would be a zero-second segment, which ffmpeg drops without an error — a shot silently missing from a cut you approved.