Make: Generate Coloring Book Line Art with AI

Click Run Rest and let it work. Every page that has a prompt and no image yet is queued and generated one by one — you’ll see each card move through Queued (3/50) → Waiting → Made. Generation is paced at roughly one image every 12 seconds to respect provider rate limits, so a 50-page book takes about 10 minutes. You can leave the app alone; an Abort button is available if you need to stop (pages already made are kept).

When the run finishes, review the grid:

• A page you don’t like? Hover it and click the edit button — see Editing a page below.
• A page failed (Error)? Click it to retry just that one.
• An empty page? Click it to generate just that one (it shows Click to generate).

Run Rest vs Run All

• Run Rest (recommended) — fills only the empty slots. Pages that are already made are never touched and never re-billed. This is the button you should reach for 99% of the time, including when resuming an interrupted run.
• Run All — regenerates every page, including finished ones. Because this re-spends credits on pages you already paid for, it first shows “Overwrite existing images?” telling you how many images exist. If you proceed, your current images are automatically backed up first (the dialog links to the backup folder), so nothing is ever silently destroyed. A Don’t ask again checkbox disables this confirmation — leave it unchecked unless you know what you’re doing.

Editing a Page

Hovering a finished page and clicking the edit button opens Edit Image with AI, which supports two different kinds of fixes:

• Refine the existing image — type what you want changed in Additional requirements (e.g., “remove the second frog, make the pond larger”). The current image is sent along as a reference, so the page keeps its composition and only the requested change is applied.
• Regenerate from a new prompt — the original sketch prompt is shown locked; click the lock to edit it. If you change the prompt (and leave the requirements field empty), the page is regenerated from scratch with your new prompt. The edit is also written back to your Plan table, so the plan stays the source of truth.

In both fields you can type # to mention a named character or object from your Image Pool — it’s inserted as {Name} and attaches that character’s reference image to the generation, keeping their appearance consistent.

Toolbar Reference

Control	What it does
Import images (icon button)	Reopens the import dialog at any time — you can mix imported and generated pages.
Cost / Quality toggle	Per-image generation quality. Cost (default) is cheaper and great for drafting; Quality costs more per image and is worth it for the final pass. Applies to generation from the moment you flip it.
Model dropdown	Which image model generates your pages. With an OpenAI key: gpt-image-2 (Latest) — best quality and supports style/character reference images — or dall-e-3 (Standard). With a Google key: Gemini 3.1 Flash Image.
Style anchor badge	Pick one style image from your Plan Image Pool to lock the whole book’s art style. Choose None to turn it off. Only appears once your Image Pool contains at least one style image.
Run Rest / Run All / Abort	Batch generation controls — see above.

A note on gpt-image-2: it requires a one-time organization verification on your OpenAI account. If your account isn’t verified, Coloring Book Engine automatically falls back to dall-e-3 (you’ll see “Switching to Standard Mode”) so your run never stalls — but treat that strictly as a stopgap. We don’t recommend producing your book on dall-e-3: its quality and capabilities are far more limited, and it can’t read reference images, so style anchors and character references are skipped (affected cards show a notice). Verify your organization and redo those pages with gpt-image-2, or use Gemini.

If Something Goes Wrong

Message	What it means / what to do
API Key Required	No AI key is configured — set one up via Connect Your AI on the Projects screen.
No Images to Generate	Nothing left to do: every page is made or currently in the queue.
Prompt is required for generation	The page you clicked has no sketch prompt — fill it in Plan (or via the edit dialog).
Generation Stopped	Two generations failed in a row, so the queue paused to protect your credits. Check the error shown (often a rate limit or provider hiccup) and press Run Rest to continue where it left off.
Organization Verification Required / Switching to Standard Mode	Your OpenAI org isn’t verified for gpt-image-2. Generation continues on dall-e-3 automatically; verify your organization to unlock gpt-image-2.
Style anchor was ignored (DALL-E 3 fallback)	Shown under pages generated during a dall-e-3 fallback: they couldn’t use your style anchor. Retry them with gpt-image-2 or Gemini for consistent style.