Publishing field guide
Use diagrams, examples, illustrations and chapter images only when they advance understanding or carry a necessary story beat.
Reviewed by Stefan Mitrović, Founder of Automateed · Updated July 16, 2026
60-second summary
Ebook images face constraints print never has: unknown screen sizes, reflowable text, file-size limits and dark-mode readers. Plan visuals that survive a phone — simple diagrams over dense infographics, images that read at 350 pixels wide, alt text for accessibility — and give each one an explicit job. In Automateed, images generate within illustrated workflows at one image credit each and remain replaceable per slot in the editor.
Real product steps
The platform generates, stores and exports images with the project, so the editorial questions — what, where, why — are the ones left for you.
Workflow map
01
Assign jobs before generating
List the places a visual genuinely helps: the framework diagram, the before/after, the worked example. Everything else is decoration that costs attention and file size.
02
Generate within the workflow, refine in the editor
Illustrated formats create images during generation. Walk the draft, keep what works, and regenerate weak slots with sharper prompts — one AI image credit per new image.
03
Prompt for screen-first simplicity
Ask for bold shapes, few elements and high contrast. A visual with twelve labeled parts dies on a phone; three-part visuals survive everywhere.
04
Test in the reflowable export
Download the EPUB and read image-heavy chapters on a phone-size screen at several font sizes. Images move with reflow — check what lands next to them.
05
Verify the PDF separately
Fixed-layout export freezes placement: check that images sit with their referencing text and print-destined files hold resolution at 100% zoom.
List the places a visual genuinely helps: the framework diagram, the before/after, the worked example. Everything else is decoration that costs attention and file size.
Illustrated formats create images during generation. Walk the draft, keep what works, and regenerate weak slots with sharper prompts — one AI image credit per new image.
Ask for bold shapes, few elements and high contrast. A visual with twelve labeled parts dies on a phone; three-part visuals survive everywhere.
Download the EPUB and read image-heavy chapters on a phone-size screen at several font sizes. Images move with reflow — check what lands next to them.
Fixed-layout export freezes placement: check that images sit with their referencing text and print-destined files hold resolution at 100% zoom.
Every step above describes the current Automateed interface — open a free preview and follow along with your own project.
Create a free previewThe full guide
The median ebook screen is a phone held at arm’s length, which rewrites design rules: line weights need doubling, label text needs 16-point equivalents, and any diagram past five elements needs splitting into a sequence. The honest test is brutal — shrink the image to 350 pixels wide and see what survives. Whatever dies at that size was never going to communicate.
In EPUB, images are boats in a river of text — font-size changes move everything, so “the chart below” may float a page away. Write image references that survive drift (“the pricing chart in this section”), keep captions attached to their images, and avoid meaning that depends on two visuals sharing a spread. Fixed-layout PDF keeps placement but gives up reflow — which is exactly the tradeoff to decide per book.
Oversized images bloat ebooks past store delivery limits and slow rendering on modest readers. The platform’s pipeline exports web-appropriate assets, but editorial restraint matters too: a hundred marginal images cost more than they add. Budget images like prose — every one included because cutting it hurt.
Many readers live in dark mode, where pure-white image backgrounds glow like flashlights. Favor visuals that tolerate inversion — transparent or neutral backgrounds, meaning carried by shape rather than color alone — and always pair images with functional alt text. Color-only encodings (red bad, green good) fail both dark screens and colorblind readers; add labels.
Decisions that change the result
Before you generate anything, write a one-paragraph brief per image slot. This is not “what the picture should look like.” It’s the decision log: the reader question, the answer the visual delivers, and the minimum detail needed to make that answer work. Example: “After the introduction of the pricing model, show how the numbers map to three moving parts; readers should be able to trace a single example from input to outcome without re-reading three paragraphs.” When later you cut prose or adjust chapter order, you can re-check the image’s job instead of guessing.
A useful brief includes: (1) the exact moment in the chapter when the image should appear, (2) the caption function (what should the caption add that the nearby text cannot), (3) what not to include (the common failure is crowding the picture with everything you considered), and (4) the readability constraints you will enforce. Set explicit limits such as “works at 350 px wide” and “no text smaller than a typical body-font size equivalent.” This makes style consistency objective, not subjective.
Ebook layouts tempt you to reuse whatever worked in print: a full-bleed photo here, a complex infographic there. In practice, ebooks reward specific visual types that translate well to reflow. If your job is comparison, use a two-column idea (even if you render it as cards) rather than a dense table. If your job is sequence, prefer a numbered mini-diagram set (step 1 → step 2 → step 3) that can be stacked vertically without losing the narrative.
If your job is evidence, decide whether a photo is truly evidence or just atmosphere. For example, a travel nonfiction chapter might include one “scene-setting” image and then rely on text for facts; that’s fine. But if your prose is already explaining the evidence, an extra decorative image repeats work. Put that effort into either: (a) a visual that clarifies a process, (b) a map/diagram that removes ambiguity, or (c) a character or location reference image that prevents reader confusion later in the chapter. This is how you avoid paying for images that are content, not comprehension.
Ebook images must look crisp at small sizes, but they also have to load quickly. For diagrams, aim for crisp edges and high contrast so the lines don’t turn into gray smudges when the image scales down. For photos, avoid overly stylized backgrounds and text overlays; when an image is readable only at large sizes, the ebook becomes a poster the reader can’t “zoom into” comfortably on every device. The safe approach is: treat diagrams as “UI elements,” and treat photos as “supporting evidence” that does not require reading tiny details.
A pragmatic workflow is to prepare one “candidate render” of each image type and check it at the scale it will be used in the ebook. If you can’t comfortably read labels on a small screen, rebuild the diagram rather than relying on the reader to zoom. The most common fix is not “increase resolution,” but “reduce the number of labels.” A diagram with fewer, larger labels is almost always a better ebook diagram.
Worked example
You’re planning a chapter in a nonfiction ebook about a three-step planning method. The chapter already contains prose that explains: (1) pick a goal, (2) map constraints, and (3) turn decisions into actions. You want images that reduce re-reading, not just decorate the page.
Brief: “When the chapter first defines the three elements, show how the goal sits at the top, constraints feed into the middle, and resulting actions fall out at the bottom. Reader question: what connects to what?” Visual type: simple flow/stack with three labeled blocks. Caption function: add a one-sentence reminder of what each block represents. Build rule: keep to three labels total; no extra icons.
Brief: “After the chapter shows a rough decision draft, include a before/after comparison to demonstrate what ‘clarified’ means. Reader question: what changes from version to version?” Visual type: two stacked cards labeled Before and After with a short bullet list inside each card. Build rule: the bullet list must be readable at small width; if it won’t be, convert the bullets into a single phrase per card and move detail back into prose.
Brief: “At the chapter’s worked example, include a trace line that follows one sample from input to outcome. Reader question: can I follow the same steps for my own situation?” Visual type: numbered steps with one highlighted path (for example, step 1 highlights a chosen goal, step 2 highlights a constraint category, step 3 highlights a resulting action). Build rule: the highlight cannot rely solely on red/green; it needs a shape or label so it remains meaningful in dark mode and for readers with color limitations.
Export the EPUB and read the chapter on a phone-sized screen at different font sizes. Confirm each image lands near the sentence that introduces it, and confirm the caption still makes sense even if the image shifts a paragraph due to reflow. Then open the fixed-layout PDF export and confirm the diagram text is legible at 100% zoom and that the image doesn’t overlap the text.
Three image slots, each tied to a specific reader question, are easier to build and easier to keep useful after edits. The key is using the chapter’s intent (“what the reader must understand next”) as the selection rule, then verifying the legibility and placement in both EPUB reflow and PDF fixed placement.
Avoidable mistakes
A common failure is duplicating what prose already tells in the same order. If the visual doesn’t add an answer that prose can’t deliver as efficiently (for example, tracing a process, clarifying a relationship, or showing a transformation), it’s likely decoration. Cut it or replace it with a visual that changes how a reader can reason.
If labels blur or lines merge at ebook scale, readers can’t use the image as intended. The repair is usually structural: fewer labels, thicker strokes, and a simplified layout that can be stacked or reduced without losing meaning.
Captions should either anchor interpretation (“this is the input/output mapping used later”) or add a constraint (“notice how the highlighted path changes”). Captions that merely restate the same information as the closest sentence waste space and attention.
Red/green distinctions often fail in dark mode and for color-vision differences. Use text labels, shapes, line styles, or patterns that remain distinguishable when colors invert or when the image is viewed on low-quality screens.
Keep the workflow moving
Quality gate
Run these checks against the actual manuscript, files and reader journey before publishing.
No repeated office scenes
Images match nearby text
Resolution fits output
Alt text explains function
Continue the exact workflow
Editorial note
This page is a practical workflow, not a promise of sales, ranking, publishing approval or a specific reader outcome. Platform rules and professional requirements should be checked at the point of use.
Questions specific to Ebook Images
Exactly as many as have jobs. Concept-heavy guides may earn a visual per section; essays and fiction may earn none. Count jobs, not pages.
The exports handle standard web formats — JPEG for photographs, PNG for diagrams needing crisp lines. Upload either; the pipeline manages the packaging.
EPUB reflows and rescales per device; PDF freezes your layout. They are different products — test both exports rather than assuming one predicts the other.
Fewer elements, thicker lines, bigger labels, higher contrast. Split anything complex into a numbered sequence of simple visuals.
Yes — a large share of ebook reading happens in it. Neutral backgrounds and shape-carried meaning keep visuals working when the page inverts.
One AI image credit per generated image, from your plan’s monthly allowance or top-ups. Uploading your own files is free.
Rarely as-is — dense infographics are poster designs. Decompose them into a sequence of single-point visuals that survive small screens.
Broken or oversized images are a common validation failure. Exporting from a managed pipeline avoids the packaging errors hand-built EPUBs hit.
Reflowable formats carry per-image descriptions to assistive readers. Write what the image does — “diagram of the three-step pricing model” — not decorative captions.
Not safely in reflowable text. Use section-based references (“the checklist in this chapter”) that stay true wherever reflow puts the image.
Name files using stable slot intent rather than temporary descriptions. For instance, use something like “ch03_goal_mapping” or “ch03_decision_before_after” so that when you regenerate or swap an asset, you know which job it belongs to. This reduces the chance of replacing the wrong image during an edit pass.
Use text-in-image only when the graphic needs it to communicate quickly (labels on diagram nodes, short step names, a small set of terms). If you need long paragraphs, prefer moving that information into prose and turning the image into a structural reference. In reflow formats, long text inside an image is likely to become unreadable at small sizes.
Explore next
Learn how to move from brief to outline, chapter drafts and human revision without accepting generic output.
Open guideUnderstand reflowable and fixed layouts, what Kindle accepts and why one master file rarely serves every channel perfectly.
Open guideLearn trim sizes, margins, bleed, fonts, front matter, EPUB checks and the difference between digital and print interiors.
Open guideUse your own topic
Review the outline, visual direction and available chapters before deciding whether to continue the full project.