In short: Yes, ChatGPT can make a short comic and keep a character fairly consistent within one conversation, but it gets harder past a few pages. ChatGPT Images follows instructions, edits images and renders short text well; you still manage references, check every panel and letter long dialogue yourself. A comic generator stores the cast once.
Last checked: October 2026 against OpenAI’s help centre, pricing page, terms and usage policies — see Sources. We make Comistitch, so weigh our side accordingly.
ChatGPT vs Comistitch for comics at a glance
| Criterion | ChatGPT (ChatGPT Images) | Comistitch |
|---|---|---|
| Image model | ChatGPT Images 2.5 (September 2026); DALL-E GPT retired August 30, 2026 | Google Gemini image models plus other models, matched to each genre and style through in-house R&D |
| Character consistency | Re-attach a reference image in one conversation; you check drift | Each character made once from a portrait reference, reused in every panel |
| Multi-panel page | Ask for a panel grid in one image | Story split into panels and laid out per page |
| Dialogue | Short quoted text works best; dense text may need fixing | Dialogue placed in speech bubbles on the page |
| Fixing one panel | Selection editor — “highlights are not always precise” | Reroll the page or a single panel |
| Free use | Limited image creation on Free | 100 credits at signup, no card (about 2–3 pages, watermarked, personal use) |
| Paid image use | More on Go, more accurate on Plus, unlimited on Pro (abuse guardrails) | Starter 12.99 USD/mo, 750 credits; a page costs 30 credits |
| Commercial use | You own the output under OpenAI’s terms | Elite, or Starter/Pro billed yearly |
| Best for | One-off strips, quick experiments | Multi-page stories with a recurring cast |
Comistitch is an independent product and is not affiliated with or endorsed by Google. Model names mentioned here belong to their respective owners.
Each criterion is explained below, followed by a step-by-step worked example in ChatGPT and the same scene in Comistitch.
What happened to the ChatGPT + DALL-E comic workflow?
DALL-E is gone from ChatGPT; every comic made there now goes through ChatGPT Images.
OpenAI retired the official DALL-E GPT on August 30, 2026 (ChatGPT release notes). Image creation moved to ChatGPT Images: version 2.0 arrived on April 21, 2026 with “images with thinking”, which plans and refines an image before drawing it, and version 2.5 followed on September 8, 2026 with sharper details, more precise editing and faster generation. ChatGPT Images is available on every plan, on web, iOS and Android (Images in ChatGPT).
So the old two-step “ChatGPT writes, DALL-E draws” pipeline is now one conversation: the same chat writes the script, draws the panels and edits them.
Can ChatGPT keep a character consistent across images?
Fairly well within one conversation and a few panels; less reliably across a long series.
OpenAI does not publish a consistency figure, and we have not run a measured test, so here is what OpenAI itself recommends (Creating images with ChatGPT):
- Revise step by step. “Step-by-step revisions help maintain consistency as you refine the image.”
- Lock what must not change. For edits, say “Change only X. Keep everything else exactly the same.”
- Use uploaded references by number. Refer to “image 1” and “image 2” when you combine a character reference with a style reference.
In practice that means one reference image of your character, attached to every panel request in the same chat, with the same written description each time. A Plus subscription adds Projects (ChatGPT pricing), which is a convenient place to keep the reference image and a character sheet between sessions.
Where it gets harder: a new conversation starts without your earlier images, two or three recurring characters multiply the checking, and props such as a specific jacket patch are easy to lose. None of this stops a short comic; it adds supervision to a long one.
Comistitch handles the cast differently. You create each character once from a portrait reference, and the builder attaches that reference to every panel the character appears in, so the face and outfit stay consistent across pages. The character consistency guide explains the reference techniques that work in any tool.
Can ChatGPT put readable dialogue in speech bubbles?
Short lines, usually yes; a page of real dialogue, not reliably.
OpenAI’s help centre says ChatGPT Images “can follow instructions to add text” (Images in ChatGPT). Its prompting guide adds that text works best when you “put text in quotes or ALL CAPS” and “keep text short”, and for dense layouts or heavy in-image text it suggests you “consider polishing in design tools if needed” (Creating images with ChatGPT).
For a comic that translates into two habits. Keep each speech bubble to a short line, and proofread every word, because a single wrong letter in a bubble is the first thing a reader notices. If your script is dialogue-heavy, generate the panels with empty bubbles or none and letter them afterwards.
Comistitch places your dialogue in bubbles as part of the page. It is not immune to lettering mistakes either, so read the page and reroll a panel if a word comes out wrong.
Can ChatGPT lay out a full comic page?
Yes, as a single image — but you cannot treat its panels as separate pieces.
You can ask ChatGPT for “a four-panel comic page, two rows of two” and it will draw the grid, in any aspect ratio you choose (Images in ChatGPT). The catch comes when panel three is wrong. You either regenerate the whole page, which can change the panels you liked, or use the editor’s selection tool, which OpenAI notes is imprecise: “Highlights are not always precise, and edits may extend beyond the area you selected.”
The alternative is one image per panel, which gives you control but moves the page layout into another app.
In Comistitch the page is the unit of work. Your story is split into panels and laid out, and you can reroll the whole page or a single panel without touching the others.
How much ChatGPT image generation do you get on free vs paid plans?
Enough to try a comic on Free; a series realistically needs a paid plan.
OpenAI’s pricing page describes image allowances in words rather than numbers:
| ChatGPT plan | Image creation |
|---|---|
| Free | Limited image creation |
| Go | More image creation (plan may include ads) |
| Plus | More complex and accurate image creation |
| Pro | Unlimited and faster image creation, subject to abuse guardrails |
Images with thinking is available on Plus, Pro and Business (Images in ChatGPT). Prices on OpenAI’s page are shown in your local currency, so check it for your country.
Comistitch is priced per page: 100 free credits at signup with no card, and a page costs 30 credits, so the free credits cover about two or three pages. Free exports carry a watermark and are for personal use. Paid plans are Starter at 12.99 USD a month with 750 credits (6.50 USD a month billed yearly), Pro at 32.99 USD and Elite at 64.99 USD (Comistitch pricing).
Can you use ChatGPT comic images commercially?
OpenAI’s terms give you ownership of what you generate, with caveats that matter for comics.
The Terms of Use (effective January 1, 2026) say you own the output and OpenAI assigns you “all our right, title, and interest, if any” in it. The same section warns that “output may not be unique and other users may receive similar output.” The usage policies rule out attempts to infringe others’ intellectual property and using a real person’s likeness without consent in ways that could confuse authenticity — so original characters only, and no look-alikes of existing comic heroes.
Comistitch includes commercial use on Elite, and on Starter and Pro when billed yearly. Free-credit exports are watermarked and for personal use.
Which should you choose?
Choose by length of story and size of cast.
- Choose ChatGPT if you want a single strip or a one-page gag and already pay for ChatGPT.
- Choose ChatGPT if you want to brainstorm the script, art direction and a rough page in one conversation.
- Choose ChatGPT if you are happy to letter long dialogue yourself in a design app.
- Choose Comistitch if your story runs several pages with the same characters.
- Choose Comistitch if you want each panel rerollable on its own without disturbing the rest of the page.
- Choose Comistitch if you want to try it free — 100 credits, no card.
Many people do both: draft the story with ChatGPT, then paste it into a comic generator. For a wider view of tools, see the best AI comic generators of 2026, and for the image-generator alternative, AI comic generator vs Midjourney.
How do you make a comic in ChatGPT step by step?
Here is a workable process for a four-panel page, using OpenAI’s own prompting advice, followed by the same scene in Comistitch.
1. Make the character reference first. Ask for one clean, full-body image of your character on a plain background, or upload your own. Save it.
2. Write the page prompt in one go. Attach the reference and spell out the grid, the beats and the exact dialogue:
Image 1 is my character Mara. Create a four-panel comic page, two rows of two,
in clean manga line art with screentone shadows. Keep Mara's face, short dark
hair with one white streak, and red bomber jacket exactly as in image 1.
Panel 1: Mara opens a small parcel in a rain-wet alley at night.
Panel 2: close-up of a glowing data chip in her palm.
Panel 3: a tall figure, Jun, appears at the alley mouth behind her.
Panel 4: Mara turns. One speech bubble per panel 3 and 4 only:
Panel 3 bubble: "THIS ISN'T WHAT THEY TOLD ME."
Panel 4 bubble: "PUT IT BACK. NOW."
No other text.3. Fix one thing at a time. If Mara’s jacket changes colour in panel 4, ask: “Change only Mara’s jacket in panel 4 to the red from image 1. Keep everything else exactly the same.”
4. Proofread and finish. Read every bubble letter by letter. If a line is wrong after two tries, regenerate with empty bubbles and letter that panel in a design app.
The same page in Comistitch. You create Mara and Jun once, each from a portrait, then write the scene as plain story text, pick a style and set one page:
Mara, a teenage courier in a red bomber jacket, has been delivering the same
sealed parcel all week. Tonight she opens it in a rain-soaked alley and finds a
data chip that is clearly stolen. Her contact Jun steps into the alley behind
her. Mara: "This isn't what they told me it was." Jun: "Put it back. Now."
Four panels, tense and low-lit, ending on Jun's face.From inside the builder that becomes one laid-out page with both characters drawn from their references and the dialogue in bubbles; you review it and reroll a panel if needed. The closest style to this example is the AI comic book maker — the free signup credits cover this page and one or two more. For turning a longer draft into pages, see turn a story into a comic with AI.
Sources
Last checked: October 2026. OpenAI updates ChatGPT often — check these pages before relying on a detail.
- OpenAI — Images in ChatGPT: ChatGPT Images 2.5, availability on all tiers, images with thinking on Plus/Pro/Business, editor limits, DALL-E GPT retired
- OpenAI — ChatGPT release notes: Images 2.0 (April 21, 2026), DALL-E GPT retirement (August 30, 2026), Images 2.5 (September 8, 2026)
- OpenAI Academy — Creating images with ChatGPT: consistency, editing and text-in-image advice
- ChatGPT pricing: image creation by plan, Projects on Plus
- OpenAI — Terms of Use: ownership of output, similarity of output
- OpenAI — Usage policies: intellectual property and likeness rules
- Comistitch pricing: free credits, plan prices, commercial use
