Writers have always painted pictures with words — but until now, that was just a metaphor.
Last week, we released a new model — 4o image generation — that’s “natively multimodal,” meaning it can think not just in text, but in images. It’s easily several orders of magnitude more capable than our previous image gen model, DALL-E 3.
- It follows complex, detailed instructions with uncanny accuracy.
- It renders text accurately and cleanly, making it easy to create typography-heavy visuals like infographics, comics, recipes, and memes.
- It preserves character consistency across multiple prompts. If you’re writing a story, your protagonist won’t randomly morph into a different person as you experiment and iterate.
- It maintains a consistent visual theme across iterations.
Art school dropout
I’ve always been hopelessly inept at visual art. Whenever I try my hand at drawing or painting, the sloppy, chaotic scrawl that congeals on the page bears little resemblance to the vivid, shimmering masterpiece soaring across my imagination.
As a writer, I’ve always relied on language as my instrument to explore ideas and convey experience. But until now, photography had been my only way to create visual work without having to rely on my own fumbling hands. And while I’m proud of my photography, a camera is tethered to reality, capturing only what I see through my lens. Through my prose, I can transport readers to scenes from my memory or imagination — but I never had a direct way to convey those images in visual form.
Last week, 4o image gen shattered that boundary. Now I can finally do what had always eluded me: summon the faded, sun-drenched snapshots in my memory and make them real for others to see.
To all my disappointed art teachers, I finally have a way to take my one true superpower — bending the English language to my will — and use it to conjure visual wonders.
How to use image gen
You create images by chatting with the model. Just describe what you want to see — and feel free to include specs like aspect ratio, color palette, and the visual style you have in mind.
Ideas for getting started
So how does image generation fit into your creative toolkit as a writer? Here are some ideas and examples from the creative community.
1. Illustrate your stories
Visualizing your story can help you imagine scenes more vividly — whether or not the final version includes graphics. Here’s an image writer Rohit Krishnan generated to illustrate a story he wrote for his then 4-year-old, about a baby triceratops and a baby T-rex on a journey to explore a volcano.
2. Visualize scenes from classic literature
Writers are also using image gen to bring literary scenes to life. Community member Dan Shipper used image gen to illustrate a scene from “About Love,” the story by Anton Chekhov.
And Wharton professor Ethan Mollick used it to illustrate a passage from T.S. Eliot’s “The Waste Land” (which he describes as “a very LLM-ish poem”).
3. Make your writing more visual
As Steven Pinker puts it, good prose is a window onto the world, allowing readers to see what you see. To make your writing visual and concrete, it can help to call forth images of the elements you’re describing — say, the bioluminescent future metropolis that looms over your sci-fi epic, the intricate mechanical contraption that rewrites your alt-history timeline, or the obscure historical figure who stars in your period rom-com.
4. Create explanatory visuals for your posts
Custom visuals can help clarify complex concepts and keep your blog posts from turning into walls of text. Journalism professor Jeremy Caplan used image gen to create infographics for his music appreciation class.
“I’m definitely using it to design illustrations for my Wonder Tools newsletter posts, experimenting with multiple styles and approaches to get professional graphic-design-style illustrations,” Jeremy told me. “I can generate visual sidebars, visual summary boxes, and other secondary visuals when I’m putting together my posts.”
5. Picture a bright future for your project
It’s an old motivational trick: visualize the finish line, the payoff for all your hard work. Use image gen to imagine your piece as a finished product: the movie billboard glowing over Times Square, the glossy hardcover showcased alongside the other bestsellers. Picture pivotal scenes as they’d appear on-screen — or hold a mental audition to cast the buzzy rising star who will play your protagonist. Making the outcome feel real won’t just fuel your motivation — it could also help your agent and editor understand your vision.
Some tips on prompting and iterating
Here are a few tips I’ve picked up playing around with the model.
-
Say exactly what you want. Be shamelessly direct about your aesthetic goals. “Visually stunning.” “Cinematic details.” You no longer need cryptic prompt engineering incantations (“8k, unreal engine, trending on ArtStation”). Just use normal human words.
-
Define your visual style. Without stylistic direction, image gen defaults to photorealism. If you want something with actual personality — a vintage 1950s illustrated children’s book with visible brush textures and expressive linework, or grainy, high-contrast black-and-white street photography — just say so.
-
Provide visual guidance. To maintain visual continuity across long chat sessions, don’t rely on text alone (“the same character as before”). Instead, upload the same reference image each time. Better yet, upload mood boards and sample images to steer the model toward what you want.
-
Expect to iterate. You might not get what you want on the first try. Just like rewriting a story draft, be prepared to adjust and refine.
-
Don’t be shy about detailed prompts. If you’ve used image gen before, you might have learned to keep prompts minimal to avoid overwhelming the model. But 4o thrives on specificity. Comprehensive prompts — even with several paragraphs of instructions — can yield good results.
Most of all, have fun. Play around. There’s something magical about using words to render a scene from your imagination into a visual reality.
What’s in your image queue?
What kinds of images are you generating? Have you discovered writerly use cases I haven’t mentioned here? Fun examples of the ones I have? Any magic prompts you swear by? Post them below! I’d love to see what you’re creating.