How to strip the slide inserts out of a lecture recording
Upload the recording and tell Vizard Agent exactly which overlay goes and which stays. Vizard Agent maps every moment a slide panel appears, inspects those frames at full resolution to find the masking boundaries, compares a clean frame to see what must be reconstructed, and removes only what you named.
What do you type into Vizard Agent?
Name both sides of the distinction. Vizard Agent finds where each overlay appears itself, but it can't guess that "the slide numbers" means the small labels and not the scene numbers that look almost identical two lines above them. Stating only the removal leaves it guessing about everything else in that corner of the frame.
Prompt
What does Vizard Agent actually do?
A lecture recording usually carries several layers of graphics at once: slide inserts, scene titles, numbering. Removing one and keeping the rest is a mapping job, not a filter, because each layer appears at different times in different places. A frame with no slide panel shows what the background should look like. In the session behind this article it:
- Built contact sheets and overview sheets across a five-minute lecture recording (1920x1080 at 30fps, 336.6 seconds) and detected the major visual changes with their timestamps.
- Reviewed all four timeline overviews to see where the layout shifted, and mapped every slide-insert moment across the timeline.
- Inspected the slide layouts and text placement at full resolution, located the panels and their labels, and read the exact masking boundaries off one full-resolution frame.
- Compared a clean talking-head frame to understand the background that needed reconstructing.
- Checked the label layout of scenes three, five, six and eight with and without the slide panel, so a mask built for one layout couldn't clip a title in another, then tested the removal on representative frames.
When does this not work well?
Removing a graphic burnt into the picture means rebuilding whatever was behind it, and that has real limits. Panels over the speaker are hardest, since there's a person to reconstruct, not a wall. Moving backgrounds defeat a static fill, and low-quality recordings show the repair. Vizard Agent tests on representative frames and shows you the actual result.
Common questions
Will there be a visible patch? Vizard Agent reconstructs from a clean frame rather than blurring, and tests it first.
What if the layout moves between scenes? Vizard Agent checks each scene separately, because one mask rarely fits a whole recording.
Does the audio change? No. Vizard Agent leaves the audio and the timing exactly as they were.