Vizard Agent

How to make a video in illustrated comic-panel style

Last updated 2026-08-16 · 5 min read

Describe the comic style you want and Vizard Agent builds the whole video as a sequence of illustrated panels. If your project already has an episode, Vizard Agent transcribes it, inspects its shots and casts a voice to match before drawing anything, so the second episode belongs to the same series as the first.

What is the short version?

Comic-panel video is a strong format because it hides the things generated footage does badly — hands, physics, continuity — behind a style where none of them apply. What it demands instead is consistency: the panels have to look like they came from one artist.

  1. Go to Vizard Agent and describe the panel style you want.
  2. Give the script, or point at the previous episode to continue a series.
  3. Say the language and the length.

What do you need before you start?

A style description and a story. Vizard Agent generates every panel, so what shapes the result is how precisely you describe the drawing — line weight, palette, era, whether panels have borders — and after the first episode it derives all of that from the last one instead.

What do you type into Vizard Agent?

Lead with the visual style and mark it as the important part. Comic-panel work goes wrong when the style instruction is buried under plot description, because Vizard Agent will faithfully deliver a well-told story that looks like something else entirely.

Prompt

Create the entire video as a sequence of illustrated comic panels. Style: [heavy line, muted palette, 1970s printing]. Story: [paste or describe]. Narrated in [Korean], [45] seconds.

Variants worth knowing:

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real comic-panel episode. Everything that happens before a single panel is drawn is about the previous episode — reading it, listening to it, looking at its shots — and that front-loading is precisely what makes the two episodes look like they belong to the same series.

  1. Reads the project record for previous work and files.
  2. Lists the workspace files for existing scripts or assets.
  3. Probes the previous video for its structure, audio and visual characteristics.
  4. Transcribes the previous video to understand its language, tone and script.
  5. Runs a shot-level inspection of the previous video's visuals.
  6. Lists the asset tools and their prices before planning.
  7. Reads the generator options for voiceover and images.
  8. Checks the voiceover help specifically for generating Korean narration.
  9. Searches for a mature, calm male voice in Korean to match what came before.

Step 9 is the one that matters for a series in a language you may not speak. Vizard Agent searched for a voice with a specific character in a specific language rather than accepting whatever the default was.

What does the result look like?

From the run this page is written from, probed on the delivered file: 1080x1920, H.264, 25fps, 45.0 seconds, AAC audio. Vertical and three quarters of a minute, with the narration mixed underneath and the panel changes cut to it rather than spaced evenly across the running time.

Ask for a wide version and Vizard Agent re-lays the panels rather than cropping them, which matters in this format because a comic panel cropped in half is no longer a panel.

Timing was not measured separately for this kind of job. The closest measured work runs a median of 28 to 38 minutes end to end, with the middle half spread considerably wider. Across all projects the median cost by tier is Flash 47, Pro 55, Max 242, Ultra 263 credits.

When does this not work well?

The style is forgiving of physics and unforgiving of inconsistency. Every panel Vizard Agent draws is generated separately, and a reader's eye notices a character's face changing between panels far faster than it notices an anatomical oddity inside any one of them.

How do you fix a result that came back wrong?

Name the panel that is wrong. Vizard Agent keeps the narration, each individual panel and the assembled video as separate pieces, so redrawing one panel leaves the reading and the timing exactly where you approved them rather than rebuilding the whole episode around the change.

How does Vizard Agent compare to doing it yourself?

The manual route is an illustrator, or a lot of prompting in an image tool followed by an edit. The illustrator gives you consistency and costs a week. The image tool gives you speed and a cast of characters who never quite look the same twice.

An image tool plus an editor Vizard Agent
Style consistency Re-prompt until it matches Derived from the previous episode
Narration A separate tool, then import Cast and generated in place
Timing panels to words By hand, per panel Built against the narration
Episode two Start the prompting again One short prompt

Common questions

Can it do speech bubbles? Yes, and it is better to ask for them as a layer than to hope they are drawn correctly inside the artwork.

Will episode two match episode one? If they are in the same project, yes. Vizard Agent transcribes and inspects the earlier episode before drawing.

Can it narrate in a language I do not speak? Yes. Vizard Agent searches for a voice with the character you describe in that language, and listens back before using it.

How many panels fit in a minute? Roughly one every three to five seconds, so a minute holds a dozen or so. Vizard Agent paces them against the narration rather than to a fixed interval.

Can I supply my own artwork? Yes. Upload the panels and Vizard Agent will time and animate them rather than generating its own.