How to show the thing you are describing instead of your face
Tell Vizard Agent to keep your recording as a small corner window and fill the rest of the frame with what you are talking about. Your audio stays as the spine, the transcript decides what each section shows, and the subtitles belong under the main picture rather than inside the little window.
What do you type into Vizard Agent?
Say the window is small and that yours is the voice. Otherwise the request reads as a standard edit with B-roll, and what comes back is your video full-frame with cutaways rather than the layout you had in mind. Concrete language helps too, because Vizard Agent builds the background from your own transcript.
Prompt
What does Vizard Agent actually do?
The transcript comes first, because what you said decides everything that appears behind you. Each section of speech gets a picture chosen for what it actually says, which is only possible once the words and their timings exist as text. In the session behind this article it:
- Checked the length and audio of a piece to camera, extracted representative frames, and confirmed exactly what was said and when.
- Checked the image, video and music generation options and their prices before anything ran, since filling minutes with generated backgrounds is the expensive part.
- Checked the background specifications and worked out how the generated images would be placed against the content.
- Generated the themed backgrounds and the piano bed in parallel, then reviewed the images together for look and composition.
- Measured the original video's level and the music's level separately, so the voice stayed in front.
- Assembled the background, the corner window and the audio into one video, then checked the frames and the waveform.
When does this not work well?
Some talks should stay full-frame. If your expression is the content, if you demonstrate something with your hands, or if the subject is too abstract to picture, the corner window throws away the reason people are watching you. Generic backgrounds read as wallpaper, and over a long video generated backgrounds get repetitive.
Common questions
How big should the window be? Vizard Agent sizes it so your expression still reads while the picture keeps the frame.
Can I supply the backgrounds? Yes. Vizard Agent places each one against the section it belongs to.
Will the images repeat? Not if you say so. Vizard Agent tracks what it has already used.