How to stack a speaker above the thing they are talking about
Tell Vizard Agent to stack the speaker above the material they are discussing. It builds a header, a speaker panel and a lower panel as measured heights rather than equal halves, keeps the speaker framed as they move, and checks the whole layout on rendered frames before cutting anything.
What do you type into Vizard Agent?
Ask for measured heights rather than a split, and say which source is the point, because the layout Vizard Agent builds follows from that. "Half and half" is easy to say and almost never right: a talking head cropped to exactly half a vertical frame loses the top of the head, and film clips squeezed into the other half lose their own shape.
Prompt
What does Vizard Agent actually do?
A stacked layout is three heights that have to add up: header, speaker, subject. Treating them as fractions of the frame produces a speaker whose forehead is cut and a panel whose contents are letterboxed inside their own box. Measuring the rendered result and adjusting is the only reliable way to land it. In the session behind this article it:
- Looked at the supplied stills and the opening frame of a director interview, and built a layout preview from the stills before touching the video.
- Re-extracted the interview windows with the speaker's face centred, measured the rendered header, gap and lower panel heights, and checked the lower panel against the speaker's crop.
- Rebuilt the header with larger type and more inset, and checked how the title and logo sat inside it.
- Cleaned the transcript, built the captions, measured loudness and reviewed a whole cutdown as a frame grid in the new layout.
- Bleeped a word and removed it from the captions at the same time, and re-extracted a clip so the interviewer's full question was included.
- Reused the established crop when a source came in at a different resolution, then uploaded the finished cutdowns.
When does this not work well?
Stacking spends vertical space on structure, so both panels end up much shorter than either would be alone. Where one of them needs the whole frame to read, such as a dense screen, a cutaway serves better than a permanent split. A speaker who moves a lot needs the crop to follow them, and two panels is the practical limit.
Common questions
Is it always half and half? No, and that is the point. Vizard Agent measures rather than splitting.
Where do captions go? In the gap, usually. Vizard Agent checks they cover neither panel.
Will the clips be distorted? No. Vizard Agent keeps their proportions inside the panel.