How to have diagrams drawn behind you as you explain
Upload your take and tell Vizard Agent you want drawings appearing behind you as you speak. Vizard Agent cuts you out of your background, tests the cutout before committing, reads your transcript for the points being made, and draws each diagram in as the words that need it arrive.
What is the short version?
An explainer holds attention when the picture is doing something. Diagrams appearing behind the speaker as each point lands give the viewer a reason to keep watching, and the timing has to come from what is being said rather than from a fixed rhythm.
- Go to Vizard Agent and upload your talking-head take.
- Say you want drawings and diagrams appearing behind you as you explain.
- Say what should be drawn, or leave it to the transcript.
What do you need before you start?
One take and a clear background situation. Vizard Agent cuts you out and replaces what is behind you, so a plain wall helps but is not required, and it reads your transcript to decide what should be drawn and when.
- The take. One camera, talking to it.
- The style. Whiteboard, notebook, dark board.
- Any specific diagrams. Or let the transcript decide.
- The frame. Portrait for social, wide for a channel.
- The pace. Tight, or left to breathe.
What do you type into Vizard Agent?
Say what appears and where it appears. Vizard Agent needs to know the drawings sit behind you rather than beside you, because that decides whether it cuts you out of your background at all, which is the most consequential technical choice in the whole job.
Prompt
Variants worth knowing:
- Handwriting typefaces. So the drawings read as drawn, not typed.
- A tightened cut underneath. Dead air removed before anything is drawn.
- Captions as well. Placed so the drawings avoid them.
What does Vizard Agent actually do?
Here is the order Vizard Agent worked in on a real explainer with progressive drawings. Two steps are the ones worth copying: it tested the background cutout on a six-second slice first, and it measured the exact rows the captions occupy before placing anything.
- Probes the uploaded video and checks the past project notes.
- Transcribes and downloads the footage, then reads the transcript structure.
- Extracts a frame grid and looks at it to see the framing and the background.
- Checks the background removal and motion graphics options and lists the effects library.
- Tests the cutout on a six-second slice rather than the whole video.
- Looks for handwriting typefaces and inspects the cutout quality closely.
- Re-checks the cutout with the correct decoder and views the test by eye.
- Measures the caption band position and locates the caption rows numerically.
- Lays out the cut list, renders the tightened cut, and draws each diagram in against the transcript's timing.
Step 5 is the habit that saves the job. Cutting yourself out of a two-minute video takes real time, and finding out afterwards that the edges are wrong means doing all of it again — six seconds tells you the same thing in a fraction of the time.
What does the result look like?
From the run this page is written from, probed on the delivered file: 1080x1340, H.264, 30fps, 136.87 seconds, AAC audio. Portrait, a little over two minutes, the speaker cut out and placed over a board on which drawings and diagrams appear as each point is made, with captions running clear of them.
Two minutes with a dozen drawings means something new appears roughly every ten seconds. That is the rhythm that holds attention: not constant motion, but a steady sense that the board is filling up as the explanation goes on.
When does this not work well?
Cutting a person out of their background is the hardest part of this format, and the conditions on the day decide how well it goes. Vizard Agent tests the cutout early precisely because these problems are cheaper to find than to fix.
- Loose hair and glasses are hard edges. The cutout shows it first.
- A busy background bleeds. Plain walls cut out far more cleanly.
- Matching clothes vanish. A green top against a green screen disappears.
- Fast gestures blur. Motion blur has no clean edge to find.
- Dense speech leaves no room. Some points arrive too fast to draw.
How do you fix a result that came back wrong?
Name the drawing or the moment that is wrong. Vizard Agent keeps the transcript, the cutout tests, the caption row measurements and the cut list, so one diagram can be redrawn or re-timed without redoing any of the background separation underneath it.
- "That diagram is wrong." Redrawn against the same transcript timing.
- "The drawing hides my captions." Repositioned from the measured caption rows.
- "My edges look rough." The cutout is retested and the settings adjusted.
How does Vizard Agent compare to doing it yourself?
By hand this means keying yourself out, drawing every diagram in a graphics tool, then animating each one to appear at the right moment. The animation timing is the slow part, because each drawing has to be matched to a sentence you have to find first.
| By hand | Vizard Agent | |
|---|---|---|
| Cutout | Key the whole video, then check | Tested on six seconds first |
| Drawings | Make each one separately | Generated against the transcript |
| Timing | Find the sentence, keyframe it | Drawn in on the words that need it |
| Captions | Hope nothing collides | Rows measured, drawings placed clear |
Common questions
Do I need a green screen? No. Vizard Agent removes an ordinary background, though a plain wall gives a cleaner edge.
How does Vizard Agent know what to draw? It reads your transcript and draws to the points being made rather than on a fixed timer.
Can I specify the diagrams myself? Yes. Vizard Agent draws what you name and works out the timing from your words.
Will the drawings look handwritten? They can. Vizard Agent looks for handwriting typefaces so the board reads as drawn.
What if my cutout looks bad? Vizard Agent finds that out on a six-second test and adjusts before processing the whole video.
Why test the cutout on a slice first? Because separating a person from their background across two minutes is slow, and discovering a bad edge at the end means doing all of it again. Six seconds answers the same question almost immediately.
Can it tighten the cut as well? Yes. Vizard Agent removed the dead air here before drawing anything in.
Will the drawings collide with my subtitles? No. Vizard Agent measures which rows the caption band occupies and keeps the drawings clear.
What frame shape works best? Portrait for social, widescreen for a channel. The board has more room in widescreen.
Can I reuse the board style on later videos? Yes. Vizard Agent keeps the background, the handwriting typefaces and the drawing style, so a series of explainers looks like one set rather than a run of separate experiments.
Does Vizard Agent check the finished explainer? Yes. It reviews the cutout and the drawing placement before delivery.