How to switch the camera to whoever is talking in a recorded panel
Upload the recording and tell Vizard Agent to cut to whoever is speaking. Vizard Agent transcribes the episode to find the speaker turns, matches each voice to the face on screen, builds a switching plan that goes full-screen for long answers and back to the grid for quick exchanges, and smooths the flicker out.
What do you type into Vizard Agent?
Describe the switching rule rather than the timestamps. Vizard Agent needs the policy, what counts as a long turn and what counts as an exchange, and it derives every cut from the transcript rather than from a list you have to write. Give it the intro music and the artwork too if the episode should open and close properly.
Prompt
What does Vizard Agent actually do?
It does what a live director does: cut to whoever is talking. That means knowing who is speaking and which box on screen they occupy. Cutting on every speaker change makes the picture flicker during a rapid exchange, so the plan is smoothed after it is built rather than applied as written. In the session behind this article it:
- Sampled frames across a 56-minute panel episode to see the layout, and transcribed it in full to find who talks when and who talks longest.
- Scanned the whole video for changes between grid and split-screen, measured the exact grid geometry, and viewed the four-person and two-person layouts at full size.
- Identified which on-screen person matched each voice, found the exact moment the layout switched, and measured the music and speech loudness three separate ways.
- Built the cut plan, full-screen for long talkers and grid for quick exchanges, then smoothed out the flickery short cuts.
- Delivered the full episode at 1920x1080, H.264, 25fps and 3374.66 seconds with AAC audio, the artwork and music fading in at the top and again at the end.
When does this not work well?
Automatic switching depends on telling the speakers apart, both in the audio and on the screen. Overlapping speech has no single speaker, similar voices are hard to separate, and a layout that moves means the boxes need re-measuring. A badly composed box looks worse full-screen, and very short turns should not cut at all.
Common questions
How does it know who is speaking? Vizard Agent transcribes the episode for speaker turns and matches each voice to a face.
Will it cut on every single change? No. Vizard Agent smooths the plan so rapid exchanges stay in the grid.
What if two people talk over each other? Vizard Agent holds on the grid rather than picking a winner between them.