Explain
A diagram makes a relationship, process or comparison faster to understand than another paragraph.
Visual publishing guide
Images for Books: Plan Visuals That Actually Help the Reader
Decide what every image must do, keep one visual language across the manuscript, and prepare artwork that survives phones, EPUB reflow and printed pages.
Quick answer
Book images earn their place by doing a job: explaining a concept, orienting the reader, or carrying a story beat. Plan images as a system — one purpose, one consistent style, correct resolution for the output — instead of decorating every chapter with stock filler. In Automateed, illustrated formats generate images inside the workflow, every image is replaceable in the editor, and cover art gets its own designer at 1600 × 2560 px.
Editorial decision
Give every image one identifiable job
If removing an image changes nothing for the reader, it is decoration. Decoration is not automatically bad, but repeating it across chapters makes a book feel padded and generated.
Orient
A map, labeled screenshot or visual example shows the reader where they are and what to notice.
Carry the story
In children’s books and visual narratives, the image contains action or emotion the prose should not repeat.
Prove
A real product, recipe or before-and-after photograph gives evidence that generic generated decoration cannot.
Inside Automateed
How to manage book images in Automateed
Image work happens in three places: the generation workflow (formats that use imagery create it automatically), the editor (replace and regenerate), and the Cover designer.
- 01
Let the format drive the image plan
Choose the right workflow first: children’s books generate page illustrations, cookbooks generate dish images, guides generate chapter visuals. The format decides where images belong; you decide what they show.
- 02
Review every generated image against its job
In the editor, walk the book and ask per image: does this explain, orient or advance? Regenerate the ones that merely decorate — each regeneration costs one AI image credit.
- 03
Keep one visual style throughout
When regenerating, repeat the same style vocabulary in each prompt (medium, palette, mood). Style drift between chapters reads as carelessness even when individual images are strong.
- 04
Replace where reality beats generation
Upload real photographs, diagrams or screenshots where authenticity matters — your product, your process, your recipe. JPEG, PNG and WebP uploads are supported.
- 05
Check resolution against the output
Screen-only books tolerate lighter images; print demands roughly 300 DPI at the printed size. Verify image sharpness in the exported PDF, not the editor.
- 06
Do the cover separately in the Cover designer
The cover is a marketing asset with its own rules — thumbnail legibility, genre signal — and its own tool, saving at 1600 × 2560 px with typography kept editable above the art.
Result: A book where every image has a stated job, one consistent style, output-appropriate resolution — and a cover engineered for the shelf rather than the interior.
300
DPI for print
350 px
phone-width test
1 job
per image
Format-by-format plan
Different books give images different responsibilities
| Book format | Best image job | Primary quality check | Editorial rule |
|---|---|---|---|
| Children’s picture book | Page illustration | Character and palette continuity | Illustration carries the story beat |
| Cookbook | Dish photograph or illustration | The image must match the written recipe | Replace with tested-dish photography where possible |
| Business or how-to guide | Diagram, screenshot, worked example | Legibility at phone width | Avoid generic people-at-laptops filler |
| Novel | Usually none | Map or restrained chapter ornament only | Spend the visual budget on the cover |
| Workbook | Example, scale, worksheet cue | Must support an action | Leave enough writing and print space |
The complete visual review
Relevance, continuity, rights and accessibility
Image planning by book format: where visuals earn their keep
Formats disagree about images. Children’s picture books are image-first — the illustration carries the beat, text supports it. Cookbooks use images as proof — the dish photo answers “will this work?” Technical guides use diagrams as compression — one flowchart replaces four paragraphs. Text-led nonfiction and novels mostly need no interior images at all, and forcing them in signals padding. Decide which category your book is in before generating a single visual.
Consistent illustration style across chapters
Readers register style drift instantly — a watercolor chapter one and a 3D-render chapter five make the book feel assembled from leftovers. Consistency comes from prompt discipline: fix the medium, palette and mood in a reusable phrase and prepend it to every image prompt. For character-led books this extends to continuity — the same character must remain recognizable, which means describing identifying features identically each time.
Image rights, AI art and marketplace rules
Images carry legal weight: confirm you hold rights to everything in the book — generated art per your platform terms, photos by license or ownership, and nothing that reproduces trademarked characters or imitates identifiable artists. Marketplaces increasingly ask how images were produced; keep production records and answer disclosure questions truthfully. For children’s content especially, review every image manually — automated generation is not an appropriateness check.
Alt text and accessible book imagery
Accessible books describe their images. Write alt text that states the image’s function — what it shows and why it is there — rather than a caption repeat. Reflowable formats carry alt text to screen readers; fixed PDFs benefit in tagged workflows. The habit also sharpens editorial judgment: an image whose function you cannot describe in one sentence probably has none.
Decisions that change the result
Turn each image into a “reader action” (not a scene)
A strong book image does something specific: it reduces confusion, confirms a sequence, or helps the reader picture an item before they meet it in the text. When you plan, write a one-line “reader action” next to each image slot. Examples: “Before Chapter 3, show the finished setup so readers can compare their progress,” or “After the definition, display the labeled parts so the next paragraph is easier to follow.” This framing prevents accidental padding, because you can tell whether the reader action is truly met by the image you choose. If you can’t name the action, pause and rework the placement or the asset type (diagram vs illustration vs real photo).
Tradeoff to understand: images that look impressive but don’t change what the reader can do are still decoration. That matters because decoration costs time during review and increases the chance of inconsistent style, lighting direction, or character appearance. Purpose-based planning keeps your editorial workload proportional to actual benefit.
Choose the right image type for the job (diagram, illustration, screenshot, photo)
Books fail when the image type doesn’t match the cognitive task. If readers need to locate items, a labeled diagram usually outperforms an illustration. If they need a sense of atmosphere or character emotion, an illustration can carry meaning that a diagram can’t. If they need to replicate an interface step, a screenshot (from your actual workflow) is more trustworthy than a generated facsimile. If they need evidence that something works, a photo of the real result is the closest match to “proof.”
A practical planning method: for each image slot, write three labels: (1) what the reader must notice, (2) whether labels are required, and (3) whether the content must be exact. Then pick the asset type. Exact-but-label-heavy usually means real screenshots or photos plus an overlay-labeled version (if your format supports it). Not-exact-but-understandable means illustration or diagram.
Prompt discipline that avoids style drift
Keeping one style is not just “use the same medium.” It’s also about repeating the same constraints so the generator doesn’t improvise. Create a reusable style prefix that includes: line weight or rendering style, background simplicity level, lighting direction, color saturation, and “no text in the image” if your layout will provide typography. Then add only the scene-specific details after that fixed prefix.
Character continuity is the hardest place style drift shows up. Don’t rely on “same person” phrasing. Instead, include stable identifiers every time: age range, hair style and color, defining clothing items, and any recurring accessory. For side characters, decide early whether they are “named and consistent” or “background and interchangeable.” That decision changes how much continuity you enforce. If you want recognizable recurrence, treat those characters like assets with consistent descriptors; if not, keep them generic to avoid awkward mismatches.
Worked example
Example: placing images in a 6-chapter how-to guide
You’re preparing a text-first how-to guide with six chapters and you want images only where they prevent mistakes. Your workflow will export to a reflowable ebook and a print-ready PDF. You already know you need diagrams more than illustrations, and you have a set of process photos from your own work.
- 01
Create the image job list per chapter
Chapter 1 (setup): one image slot labeled “reader action: confirm you’ve gathered the right components.” Chapter 2 (first steps): one slot labeled “reader action: show the starting configuration before step-by-step instructions.” Chapter 3 (common mistake): one slot labeled “reader action: demonstrate the correct direction so readers avoid the most frequent error.” Chapter 4 (sequence): two slots labeled “reader action: show step order” (A then B), not separate scenes. Chapter 5 (troubleshooting): one slot labeled “reader action: map symptoms to checks” (a flowchart-like diagram). Chapter 6 (final result): one slot labeled “reader action: compare your output to the expected end state.”
- 02
Choose real assets where authenticity reduces risk
For the setup and final result slots, you upload your own photos because the reader must recognize your exact components and finish. For the troubleshooting checks and the symptom-to-check mapping, you plan a clean diagram style. For the first configuration and the sequence steps, you use generated illustrations that match your diagram style—because you’re teaching structure and order, not documenting a specific device model.
- 03
Lock a reusable visual system before generating
You decide on a consistent diagram look: flat vector-like rendering, white/soft neutral background, limited palette, no embedded text inside the images, and high-contrast arrows/labels supplied by the surrounding layout where needed. You reuse that system phrase in every generated image prompt, then add only the part that changes (configuration A vs B, which arrow direction, which element is highlighted).
- 04
Check relevance in the editor using a per-image test
You scroll chapter by chapter and apply a simple test: “If a reader skipped this image, would they misunderstand the next paragraph?” If yes, the image earns its place; if no, you either remove it or change it to a stronger job type (often a diagram). Then you regenerate only the image that fails the test. Remember that regenerations are the fastest way to burn time and credits when you’re not decisive, so you audit before you regenerate.
At the end, your guide uses 7 images total, but each has a defined job: confirm readiness, show starting setup, prevent the top mistake, clarify order, map troubleshooting logic, and anchor the expected outcome. The result reads as intentionally illustrated instead of decoration-heavy.
Avoidable mistakes
What usually breaks this workflow
Using multiple image styles because “they’re different scenes”
Different scenes are expected to differ. Different render styles are not. If a reader can point to where the artwork changes between chapters, you’ll feel the impact as “carelessness,” even if every single image is technically good.
Choosing illustration when a labeled diagram is required
If readers must locate parts or follow a direction, a diagram with clear structure and consistent arrows usually outperforms a decorative illustration. When you pick the wrong type, the image looks relevant but doesn’t actually reduce confusion.
Letting resolution be judged by the editor preview
Editor previews can mask softness that becomes obvious in print exports. Verify sharpness in the exported PDF at a zoom level that matches how the printed page will feel to a reader.
Including embedded text inside generated images when you plan to edit layout typography
When text inside the image needs later changes (chapter titles, label wording, formatting), embedded text locks you into re-rendering. Prefer “no text inside” for generated assets when your layout will provide the typography.
Images for books FAQ
Questions authors ask before export
Which book formats generate images automatically?
Illustrated workflows — children’s books, cookbooks, storybooks, travel and visual guides — create images as part of generation. Text-led formats stay text-led by design.
What does regenerating an image cost?
One AI image credit per generation, the same unit the cover background uses. Plans include monthly image credits and top-ups are available.
How do I keep characters consistent across illustrations?
Describe identifying features identically in every prompt — age, hair, clothing, palette — and regenerate until continuity holds. Consistency language is as important as scene language.
Can I use my own photos instead?
Yes — upload JPEG, PNG or WebP wherever an image slot exists. Real photography beats generation for products, processes, food you actually cooked and anything evidential.
What resolution do print images need?
About 300 DPI at printed dimensions. Export the print PDF and inspect images at 100% zoom — screen-fine images are the classic print casualty.
Do novels need interior images?
Almost never beyond maps or chapter ornaments. Fiction’s imagery budget belongs to the cover, where it does the selling.
How many images should a nonfiction guide include?
As many as have jobs — typically one orienting visual per major concept. Ten purposeful diagrams beat forty decorative stock scenes.
Who owns AI-generated book images?
Usage rights follow the generating platform’s terms and your jurisdiction. Keep records of what was generated versus uploaded — the same records marketplace disclosure questions ask about.
What is alt text and do ebooks need it?
A short functional description of the image for readers using assistive technology. Reflowable formats support it and inclusive publishing expects it.
Why does my cover need a separate tool?
Because covers are retail assets judged at thumbnail size with genre-specific typography — different rules, different dimensions (1600 × 2560 px), different tool.
How should I label image purpose during review so I don’t lose track of what each one is for?
Use a consistent internal tag you paste next to each image slot in your workflow—something like “Explain,” “Orient,” or “Advance,” plus a one-sentence reader action. During editing, you can then justify removal or regeneration without rereading the entire manuscript to remember intent.
What’s a good rule for deciding between “one broad image” vs “several small images” in the same chapter?
If the reader’s job is to understand a single concept or relationship, one broad image is usually cleaner. If the reader’s job is to follow multiple discrete steps that must be compared, break it into a small sequence (A then B, or Step 1 then Step 2). Avoid splitting when the steps are just repeating visual setup without new information.
Visual quality gate
Create the book first. Keep only the images that earn the page.
Generate a free preview, review the visual direction and replace generic scenes before the final export.