Let the AI Read Your Comics, Not Draw Them

A note for my friends from the comic seminar: This piece is the practical extension of my mini-workshop at the Erlangen Comic Seminar in 2023 and the article “AI: Your Enemy and Helper” that grew out of it (both in German). The “and” was deliberate; it was never an either/or for me. The really big questions I’m leaving out on purpose: technocracy, the “Dark Enlightenment”, the military use of AI in Ukraine, Gaza and elsewhere. Those deserve more tears, with good reason, than the fate of our own draftsmanship, and in our artistic navel-gazing they barely come up at all. But they belong at a different table. This is about craft, about what you can try tomorrow.

“AI: friend or foe?” That was the framing on the panel, and honestly I found the question flat before it was even discussed. Not because the worry is groundless. The fear of losing work is real, generative style-theft is a genuine problem, and the copyright questions are anything but settled. That fight has to happen, and in many places it is being fought hard, rightly so.

But the debate has fixated on a single use: the AI that draws (animates, composes, and so on). The machine that imitates a style, spits out an image, is meant to replace a human at the drawing table. Next to it sits a second register almost nobody talks about: the AI that reads.

An AI that doesn’t generate a picture but looks at one. That doesn’t copy my style but tells me what lands with it. That is a very different lever, and a far less threatening one.

We touched on this on the panel with the example of translation: AI can do an enormous amount here without making translators and editors redundant. Their role shifts toward curation, co-creation, polish instead of raw conversion. Not replaced, but promoted into a new job profile, if you let it happen.

The same pattern carries over to a lot of work steps, as soon as you think of AI as receptive rather than productive. And the best way into this field isn’t the most comfortable one, it’s the one almost nobody argues against.

For blind and visually impaired people, comics have always been a wall: a medium that tells almost everything through the image can’t be unlocked with a screen reader. This is exactly where a reading AI is strong.

It can draft a panel-by-panel image description (who’s in the frame, what’s happening, what the line and the colors carry) and prepare the most laborious layer of an accessible edition. From that description text you can then produce what was previously too expensive to happen at all:

• an audio track that makes the comic experienceable like a little radio play: dialogue plus a described visual layer, read aloud;

• the text basis for a Braille or tactile edition, where the visual storytelling is translated into descriptive language;

• a version in plain language for people with cognitive impairments or reading difficulties.

Important: the AI delivers the draft, not the finished product. The artist decides whether “the figure seems threatened” is right or whether more ambivalence belongs there. That turns accessibility from a duty tacked on at the very end into something doable. And nobody loses a job in the process, because this work simply wasn’t being done before. Here AI adds something that would otherwise have been missing. A better way into the whole debate is hard to imagine.

Closely related, and already mentioned on the panel: the same reading logic carries a comic across language borders. An AI can draft a raw translation of the speech balloons, onto which a human then sets the actual tone, the wordplay, the rhythm, the lettering. Combined with the description text from above, a multilingual audio version emerges almost as a byproduct. Here too the translator isn’t replaced. She’s freed from the mechanical first pass and can concentrate on what only she can do: the feel for voice and culture.

Graphic novels carry a lot of thankless diligence in things nobody enjoys doing and that still get noticed when they’re missing. An AI that has read the whole work can assist precisely here:

• Continuity: Does the character suddenly have no scar on page 12? Is the jacket a different color in chapter 5 than in chapter 2? Is the order of the flashbacks right?

• Linguistic consistency: Does a character sound the same across 200 pages, or does their tone drift without anyone noticing?

• Lettering and errors: inconsistent spellings, forgotten quotation marks, speech balloons set too tight.

The editorial judgment, what’s intentional and what’s an error, stays with the human. What the AI takes off your plate is the dull combing-through. That shifts editing from hunting for mistakes to curating: away from searching, toward deciding.

Comics are often surprisingly research-heavy, historical and documentary material especially. If you’re telling a story set in Berlin in 1925, you need to know what a tram looked like, what uniform the conductor wore, how a shop sign was lettered, what hung at a newspaper kiosk, how people dressed. Here an AI is a fast first port of call: it offers leads, search terms, and threads to pull on.

The catch, and you owe your colleagues this honestly: AI occasionally invents details that sound plausible and are simply wrong. It’s the first address, not the last. What it gives you, you check against the archive, the photo book, or the expert. Excellent as a research accelerator, useless as a source of truth.

Plenty of people are brilliant at the drawing table and suffer the moment they have to explain their own work: the blurb, the crowdfunding pitch, the grant application, the three sentences for a jury. An AI that has read the comic is a good sparring partner for putting your own idea into words. Not by inventing something, but by reflecting back what the thing is apparently about and offering phrasings you accept, reject, or rewrite. In the German funding jungle especially, that has gotten more than one good project over the application hurdle in the first place.

If you’ve been drawing for twenty years, you’re sitting on a mountain of your own work that nobody can survey anymore. An AI can make that jumble navigable: descriptions, tags, a searchable index. “Show me every page where it rains” or “which stories take place at night” becomes answerable. It touches no one else’s copyright, because it’s purely your own material. And it’s useful in an unspectacular way.

And finally, maybe the loveliest case: the AI as a study partner working through other people’s masterpieces. “Explain how Moebius handles negative space.” “What makes the panel timing on this page so breathless?” “Why does this page break work and the one before it doesn’t?” You don’t learn here by having the machine draw, but by having it help you see more precisely. And that is exactly the receptive stance this whole piece is about.

In every one of these cases the decisive thing stays with the human: the judgment, the style, the authorship. The machine reads, checks, describes, organizes. It doesn’t replace your eye, it gives you a second one.

None of this makes the tools innocent. They come from the same corporations whose handling of training data is rightly being fought over. But the question of which tool is the smaller one. The bigger one is: which use shifts power toward creators, and which shifts it away from them?

An AI that copies my style subtracts value. An AI that makes my comic readable for more people, or gives me an honest first round of feedback, adds some. That’s not an argument for the corporations, it’s an argument for not letting them dictate the use.

The most convincing version of all this isn’t an argument anyway, it’s an experience. So here are four prompts to try: copy one, upload a comic of your own as a PNG or PDF, paste the prompt in, and let it read.

Each prompt sets up a different reading stance. They don’t replace human editing or a colleague you trust. But they give you four points of view before you even ask anyone: free, instant, at any hour.

Reads your comic the way a real audience would, seeing it for the first time. It doesn’t judge whether it’s “good”, it reports what actually happens in a reader’s head. Swap the target group in brackets for your own: fantasy fans, comics nerds, a school class, whoever.

You are reading the uploaded comic for the first time, as [TARGET GROUP: e.g. a
general audience with no comics background / role-playing and fantasy fans /
lifelong comics nerds]. You know neither me nor my intentions. Don't rate anything
as "good" or "bad"; instead, report honestly on your reading experience:
1. What do you think happened here? Retell the story in your own words.
2. Where did you stumble, flip back, or read something twice?
3. What stayed unclear to you?
4. What mood or emotion came up while reading, and exactly where?
5. What stays with you a day later?
Answer in the first person, as this reader, short and concrete.

An experienced, well-meaning colleague who sees the craft. Strengths first, then concrete, actionable improvements, and at the end the one question that moves the revision furthest.

You are an experienced comics artist giving me collegial, constructive feedback on
the uploaded comic. You want the work to get better, not to make me feel bad. Be
concrete, not general.
1. Strengths: What works really well, craft-wise? Name 2 to 3 specific spots.
2. Storytelling and rhythm: Is the panel timing right? Where is the pace too fast,
   where too slow?
3. Page architecture and flow: Does the composition lead the eye where it should go?
4. Readability: Do the lettering, speech balloons, and text-to-image balance work?
5. Improvements: 2 to 3 concrete, actionable suggestions.
6. One question: Ask me the single question that will help me most in revision.
Keep it short and speak to me as an equal.

No diplomacy. Names the weakest spot, the conventional, the convenient: unsparing, but in service of the work, not of cruelty. Twist the blunt spoon in the wound if you like, but tell me at the end what would save the piece.

You are an unsparing but competent critic. Politeness won't help me here. Tell me
the uncomfortable things about the uploaded comic that a nice person would keep
quiet, but in service of the work, not of cruelty.
1. The weakest spot: Where is the comic conventional, convenient, or predictable?
2. Where do I lose the reader, and exactly why?
3. What only pretends to have an idea?
4. Which one sentence do I not want to hear but need to?
No softening, no list of strengths. Twist the blunt spoon in the wound if you like,
but end with one sentence on what would save the piece.

Not a stranger, but you: older, further along, looking at your own work from a distance. What will still make you proud, what will make you cringe, and which seed is here worth developing?

You are me, the same artist, only five or ten years older and further along. You are
looking back at the uploaded comic that you made years ago. You are neither harsh
nor gentle, just honest with distance.
1. What will you still be proud of in five years?
2. What will make you cringe a little, and why?
3. What do you not see yet today that becomes obvious with distance?
4. Which seed in this work is worth developing further?
Speak to me the way you'd speak to your younger self: direct, warm, no coddling.

The point isn’t that these four voices are the right ones. The point is that you can build the voices yourself, and that the interesting question becomes a different one. No longer “friend or foe”, but:

Which reading stance would I love to have at my table that I could never otherwise get?

An editor who never tires. A child who sugarcoats nothing. A stranger from the target audience I’d otherwise never meet. My own self ten years from now. These aren’t replacement people, they’re rehearsal stages before the real people come into play.

Bring your own comics, change the prompts, write your own. And let’s talk about which voices you’re still missing.

The four voices above are readers. This one is something else: a sparring partner for a moment we all know right now. You’re sitting at the comic salon, half drenched in sweat, across from a publisher, you have two minutes, and afterward you can barely say what you actually told them.

An AI that has read your pages can help you put into words, beforehand, what it’s about. Not by inventing anything, but by building a pitch out of what’s actually on the pages, one you can say out loud. Upload a few pages as PNG or PDF (three are enough, ten work too) and use this prompt:

You are reading the uploaded comic pages and helping me build a publisher pitch out
of them. You invent nothing; you work only with what's on the pages and in my notes.
My notes:
- Working title: [...]
- Genre / target group: [...]
- What I think is special about it: [...]
- Comparable works (for context): [...]
- Stage: [e.g. finished / outline / 12 of a planned 120 pages]
- What I want from the publisher: [e.g. publication, advance, feedback]
Help me with:
1. The logline: what it's about, in a single sentence I can take into the elevator.
2. Three lengths: one sentence / three sentences / a spoken 30-second version.
3. The hook: what sets this apart from everything else landing on this publisher's
   desk today?
4. What the images deliver: what do these specific pages show that I should point to
   in the conversation?
5. Prep: the three questions a publisher will likely ask, and one honest bullet each
   on how I answer.
Write it so I can say it out loud without sounding like I'm reading. No filler, no
marketing-speak.

It replaces neither your instinct for the right publisher nor the nerve to sit down. But it takes away the dread that the words will desert you at the decisive moment. And it’s a nice proof, in passing, of this whole piece’s thesis: the AI doesn’t draw your comic, it helps you talk about it.

License

This text by Georg Graf von Westphalen is published under a Creative Commons license: Attribution, No Derivatives 4.0 International (CC BY-ND 4.0). You may share and redistribute it freely, including in commercial contexts, as long as you credit the author, link back to the original, and pass it on unchanged. License: https://creativecommons.org/licenses/by-nd/4.0/

Exception: The five prompts in this article are expressly released for copying, adapting, and reuse, including in modified form.