How to get a clean speaker crop out of a streaming platform layout
Tell Vizard Agent which person's box you want out of the recording. It finds the webcam tiles by their border colour, crops one, looks at the result to check whether the platform's frame is still visible, and then crops inside the border until what is left is only the speaker.
What do you type into Vizard Agent?
Say which person, not which coordinates. Layouts move, since a guest joining or a screen share starting will reshuffle the boxes, so an instruction phrased around a person survives the recording changing shape and one phrased around a rectangle does not. Name the output frame too, which is usually vertical for clips.
Prompt
What does Vizard Agent actually do?
A recording from a streaming studio is a picture of a layout: boxes, borders, name bars, a logo. A crop derived from measured coordinates is nearly always a few pixels generous, so a sliver of border survives at one edge. Vizard Agent works in a loop instead: crop, look at what came out, then crop tighter until the platform is gone. In the session behind this article it:
- Inspected a podcast recorded in a streaming studio, extracted frames across it to see the layout, then took a full-size frame to get the host's box coordinates.
- Detected the gold-bordered speaker boxes by segmenting on the border colour, so no template for the studio was needed.
- Cropped the host's webcam area, looked at the result to see whether the platform frame survived, cropped inside the border and looked again to confirm none remained.
- Checked segments where the content changed, concatenated the pieces losslessly, verified the streams and confirmed the audio survived.
- Rendered the clips in the requested layout and extracted timeline frames to verify subtitles, layout and overlays.
When does this not work well?
The method assumes each person occupies their own rectangle in the frame. Layouts that overlap, blur or animate the tiles never give Vizard Agent a clean box to take, and a shared screen filling the frame puts nobody in a box. Very small tiles leave too few pixels, and a layout that changes mid-recording needs a crop per section.
Common questions
Does it need to know the platform? No. Vizard Agent finds the boxes by looking rather than by template.
Can it do both speakers? Yes. Vizard Agent produces each as its own clean source.
Why not just crop to the coordinates? Because a measured rectangle is generous by a few pixels, and those pixels are the platform's border at full screen.