How to make a comedy micro-drama where only one character speaks
Give Vizard Agent the script and say which character is the only audible one. Vizard Agent auditions voices for that character, generates every other character's reactions as silent performance, builds the on-screen interface as clean vector graphics, and cuts the timing to the spoken lines.
What do you type into Vizard Agent?
Write the rule as a rule. Vizard Agent treats "she is the only audible speaking character" as a hard constraint on the whole build, which changes how the other characters are generated and how the beats are timed around them. Add a reference image, because it conveys the characters and the layout faster than a description.
Prompt
What does Vizard Agent actually do?
A one-voice micro-drama has a real constraint: one character talks and everyone else reacts. That is what makes it funny, and it means the silent characters need generated expressions precise enough to carry a beat without a line. Vizard Agent auditions the voice first, and draws any app interface as vectors, because a generated screenshot reads as fake. In the session behind this article it:
- Opened the reference image, inspected the layout dimensions, sliced it into panels and cropped the character faces out of them.
- Searched for candidate voices, tested them and listened back before choosing, then generated the speaking character's lines and measured the total duration.
- Generated the silent character's performance (the guilty look, the gesture, the suppressed laugh), reviewed each batch as tiles, and made clean portraits and key close-up frames.
- Tested the split-frame layout and generated a full interface frame, then rebuilt the interface as vector graphics when the generated version did not hold up.
- Measured the exact voiceover timings, sourced comedy sound effects for the beats, measured each audio stem's loudness, and rendered 60.98 seconds at 1080x1920 and 30fps, letting the voice set the length.
When does this not work well?
Generated characters and comic timing are each hard, and this format needs both at once. Vizard Agent generates clean reference portraits to hold a look, but long series still drift. A beat that works in your head needs to be written as a beat. Real people's likenesses and replicas of a real app's branding are off limits. And one voice carries everything: if the audition doesn't land, the episode doesn't either.
Common questions
Why only one speaking character? It's a format convention, and it makes the silent reactions do the comedy.
Can it keep the character across episodes? Yes, from the same reference. Vizard Agent generates clean portraits for that, though long series still drift.
How long should an episode be? Forty-five to sixty seconds. Vizard Agent lets the voice set the exact length.