Textbook Illustration Maker Guide: Consistent Figures Across Chapters
A practical guide for science and medical textbook authors: how to keep illustrations consistent across chapters, what to teach with diagrams vs. text, and AI prompts that hold a style sheet.
A textbook author redrew the mitosis figure five times before publication. Not because the science was wrong — it was right every time. The publisher's editor kept rejecting it because the cell color did not match the figure two pages earlier, the arrow style was different from the meiosis figure in chapter 12, and a student would have to learn the visual language twice.
That is the real cost of textbook illustration. A journal figure stands alone; a textbook figure has to live next to twenty other figures, in three editions, across two languages, and still teach the same concept the same way. AI lowers the cost of drawing — it does not lower the cost of consistency. This guide is about how to keep both under control.
Treating each figure like a journal figure. A journal figure proves a result; a textbook figure teaches a concept. Density that works for a reviewer overwhelms a sophomore.
More Posts
Create your scientific figures with AI
Thousands of researchers use SciDraw AI to make publication-ready figures for papers, grants, and journal submissions — in minutes, with no design skills.
No style sheet across chapters. Cell A in chapter 3 is blue, in chapter 9 it's green — the student thinks they are different cells. Pick the palette and icon set once, then stick to it.
Cramming the definition into the image. Long labels turn the figure into a glossary. Definitions belong in the surrounding paragraph or the caption.
Generic AI cartoons that ignore curriculum level. A figure that is fine for a high-school biology unit is too simple for a graduate cell biology chapter. Specify the audience grade level in the prompt.
No editable export. Translated editions need to swap labels — German, Spanish, simplified Chinese — and that is only possible if the figure is SVG, not flattened raster.
Too short — produces a generic, off-curriculum mitosis cartoon:
Make a textbook illustration of mitosis.
Restructured — produces a figure that fits the chapter:
Create a four-panel textbook illustration of mitosis stages for an undergraduate cell biology chapter.Panel 1 (prophase): chromosomes condense, nuclear envelope intact.Panel 2 (metaphase): chromosomes aligned at the metaphase plate, mitotic spindle visible.Panel 3 (anaphase): sister chromatids pulled to opposite poles.Panel 4 (telophase): two nuclear envelopes reforming, cell beginning to cleave.Use a teal cytoplasm and a warm pink for chromosomes — match our chapter palette.Same cell outline style as our meiosis figure: 2 px stroke, rounded oval.Labels: short single-word callouts ("chromosome", "spindle", "centrosome"). Definitions stay in the caption, not the image.Output as editable SVG so labels can be swapped for the German and Spanish editions.
The structured prompt locks in the style sheet and the teaching density — the two things that make textbook figures consistent.
Palette: 4–6 colors used across all figures (cell types, compartments, molecules).
Stroke and line weights: cell outlines, arrows, callout leaders.
Label style: font, size, abbreviated vs. full names, single-word callouts vs. short phrases.
Icon system: how a cell, organ, instrument, or molecule is drawn each time.
Caption convention: where definitions live, how long captions can be, terminology rules.
Put this in a one-page reference and paste it into every prompt. Without it, the AI will invent a new style sheet for every figure, and your editor will reject all of them.
What to notice: four panels of equal weight; labels are single words; the supporting detail (definitions, mechanism) stays out of the image; the cell color and stroke are identical across the four panels so students compare stages, not visual styles.
Create a textbook illustration explaining [process] for [grade level / course] students.Show [stage 1], [stage 2], [stage 3], and [stage 4] in equally weighted panels with matching aspect ratio.Use the palette and icon system from our style sheet: [paste style sheet excerpt].Labels: single-word callouts only. No definitions in the image — those go in the caption.Movement arrows between stages: thin, single direction, no decorative curves.White background. Editable SVG output.
Create a medical textbook illustration for [condition or organ system] for medical students.Show four states: [anatomical normal], [early disease change], [advanced disease change], and [clinical or radiological correlate].Use the same anatomical viewpoint across all four states so students compare changes, not perspectives.Labels: abbreviated anatomical names. Mark pathology with a dashed outline, not red — color blindness compliance.Style: educational atlas, no graphic content, calm clinical palette from style sheet.
Design a textbook figure comparing [concept A] and [concept B] for [grade level] students.Use two columns of equal width with identical row structure: [mechanism], [structure], [function], [example].Matching icons on left and right within each row, so the difference is visible at a glance.Short labels, definitions in caption.Use the chapter palette; A in [color 1], B in [color 2]. SVG output.
Create a labeled anatomy illustration of [organ or system] for [grade level / course] students.View: [sagittal / coronal / axial / 3-D oblique], specify a single perspective.Label the following structures: [list every structure to label]. Do not label other structures.Use leader lines, not arrows. Place labels outside the body outline so the anatomy stays readable.Style: educational atlas, matching the rest of our anatomy figures (palette and stroke from style sheet).Output as layered SVG so labels can be translated for international editions.
First-time textbook author: write the style sheet before you write the prompts. Most rejected figures come from missing style sheets, not bad prompts.
Editor or production lead: enforce the style sheet at intake. If an incoming figure does not paste the style sheet into the prompt, send it back before reviewing the science.
Co-author handling international editions: insist on SVG output for every figure. Translation is a label swap if you have SVG, and a full redraw if you don't.
Curriculum designer / instructional designer: use template 1 for processes and template 3 for compare-and-contrast. These two templates cover most learning-objective figures.
Medical illustrator working with AI drafts: use template 2 to get the layout, then redraw the anatomy yourself for the final version. AI is fast for composition, unreliable for clinical accuracy.
A journal figure proves a result for a specialist. A textbook figure teaches a concept to a non-specialist who will read another 19 figures in the same chapter. Density, label style, and visual consistency requirements are all stricter for textbook figures.
Use single-word callouts for objects students need to identify. Push definitions and mechanism descriptions into the caption or body text. A figure full of definitions stops working as a figure and becomes a glossary.
Yes — if you exported SVG and the labels live on a dedicated text layer. Shorten labels for slides, hide definitions for handouts, swap language for translated editions. All three are label swaps, not regenerations.
Textbook illustration prioritizes pedagogical clarity and cross-chapter consistency. General book illustration also includes chapter openers, decorative spreads, and editorial visuals where consistency is less critical.
Paste your style sheet into every prompt. Models default to maximum diversity unless you constrain palette, stroke, and icon style explicitly. "Match our chapter style" with no specification means nothing to the model.
For early drafts, yes. For final medical textbook anatomy, a qualified medical illustrator or anatomist must review. AI confidently draws clinically incorrect details; the editorial process must catch them before press.
Davie Chen is a researcher at the Faculty of Animation and Intermedia, University of Arts in Poznan, studying generative AI for scientific figure creation, patent illustration, and manuscript drafting. SciDraw AI is one of the research-to-product tools built from this work.
