Stage 5: Illustration
Prompts: Visual Brief (figure and diagram specification) | COMAC-Visual (visual accessibility check) | Caption Generator
Stage 5 applies to chapters with figures, diagrams, photographs, or any visual element.
For text-only chapters, skip this stage.
World model equivalent: In a world model project, visual elements are specified in
the architecture doc (Doc 03) alongside the chapter outline — what goes where, what it
illustrates, and what caption it needs. Stage 5 does this inline per chapter.
Alias coverage: IID, EIID, and IG are included below as compact alias prompts
that map directly to this stage's operational blocks.
Use when: Planning which visual elements a chapter needs before writing (or after, to spec visuals for an existing draft).
Generate visual element specifications for this chapter.
Chapter: "[CHAPTER TITLE]"
Chapter intent: [what reader should be able to do after reading]
Target reader: [knowledge level]
For each concept in this chapter that would benefit from
a visual:
1. VISUAL TYPE: [diagram / photograph / chart / screenshot /
step-sequence / comparison table]
2. WHAT IT SHOWS: [specific content — not "a diagram of X"
but "a diagram showing how ISO, shutter speed, and aperture
interact in the exposure triangle, with values for a
typical night sky shot at ISO 3200"]
3. WHERE IT APPEARS: [after which paragraph or section]
4. CAPTION: [one sentence — what the reader should notice]
5. ALTERNATIVE TEXT: [for accessibility — what is depicted
for a reader who cannot see the image]
Also: flag any concept where a visual would replace 200+
words of explanation more effectively than prose.
COMAC-Visual — Visual Accessibility Check
Use when: After adding visuals to a chapter — verify they are accessible and necessary.
Review the visual elements in this chapter.
For each figure, diagram, or image in this chapter:
[DESCRIBE EACH VISUAL — what it shows, its caption,
its position in the chapter]
Check for:
1. NECESSITY: does this visual replace explanation that
prose cannot do as well, or is it decorative?
2. SELF-CONTAINED: can the visual be understood without
reading surrounding text? (It should be.)
3. CAPTION ACCURACY: does the caption direct the reader
to what they should notice?
4. ACCESSIBILITY: is there sufficient alt-text or
descriptive prose for readers who cannot see the image?
5. INTRODUCTION: is the visual introduced in the preceding
paragraph? (Never present a figure without a lead sentence.)
For each: [KEEP / REVISE / REMOVE] — [reason]
Caption Generator
Use when: You have a visual but need a concise, accurate caption.
Write a caption for this visual element.
Visual type: [diagram / photograph / chart / screenshot]
Visual content: [describe specifically what is depicted —
include any labels, values, or data points visible]
Chapter context: [what the surrounding text is discussing]
Target reader: [knowledge level]
Generate:
1. SHORT CAPTION (1 sentence, <20 words): what the visual
shows and what the reader should notice
2. EXTENDED CAPTION (2-3 sentences, optional): if the visual
requires context to interpret correctly, add 1-2 sentences
of explanation after the core caption.
Do not begin the caption with "This figure shows..." or
"The diagram above..." — start with the specific content.
Alias Prompts (IID / EIID / IG)
These blocks provide compact alias names for the same Stage 5 logic.
IID — Illustration Identification and Description
Act as IID.
Input:
- Chapter text or chapter summary
- Target reader level
Task:
Identify every concept that should be visualized and produce,
for each visual candidate:
1. VISUAL TYPE
2. WHAT IT SHOWS
3. WHERE IT APPEARS
4. CAPTION (short)
5. ALT TEXT
Output:
- VISUAL_BRIEF_TABLE
- PRIORITY_ORDER (high/medium/low)
Execution note:
This is functionally equivalent to "Visual Brief — Figure Specification".
EIID — Extended Illustration Identification and Description
Act as EIID.
Input:
- IID output
- Chapter context paragraph for each visual
Task:
Expand each visual brief with:
1. LAYOUT guidance (size/position)
2. DIFFERENTIATION guidance (how to avoid visual overlap)
3. ACCESSIBILITY constraints
4. CAPTION QUALITY checks
Output:
- EXTENDED_VISUAL_BRIEFS
- REVISION_FLAGS
Execution note:
This combines Visual Brief expansion + COMAC-Visual checks.
IG — Illustration Generation
Act as IG.
Input:
- Final visual brief
- Target style constraints
Task:
Generate production-ready visual instructions and captions.
For each visual, return:
1. GENERATION PROMPT or DESIGN BRIEF
2. SHORT CAPTION
3. EXTENDED CAPTION (if needed)
4. ALT TEXT
Output:
- VISUAL_GENERATION_PACK
Execution note:
This maps to Caption Generator plus production handoff instructions.