Character Model Sheet

A character model sheet is a single generated image that shows a character's complete design at once — not one pose, but a full-body turnaround (front, 3/4, side, back), a head & detail close-up sheet, a color palette, and a short overview block naming the character, its role, and its personality. It's the format riggers, illustrators, and game/animation studios use to lock a character's design before it's animated, modeled, or drawn by a second artist — the point is consistency across angles, not a single nice pose.

This is a sibling to brand-identity-sheet (same two-artifact workflow, same Ideogram generation pattern) but for characters instead of brand systems, and a sibling to logo-prompting's "single mark" discipline scaled up to "single character, many views."

The two artifacts this skill produces

1. A generation prompt — a tight descriptive paragraph, Ideogram-legible, that produces the sheet in one shot.
2. A compositional deconstruction — a structured JSON breakdown of the resulting image (or of a reference image the user supplies) with a high_level_description and an elements array of {type, bbox, desc} / {type: "text", bbox, text, desc} entries. This is what makes the sheet reusable: another agent (an animator, a modeler, a second illustrator) can read the bbox spec and know exactly which panel is the front view, which is the face close-up, which is the color swatch bar, without re-looking at the image. See references/composition-spec-format.md for the full schema — it's the same schema brand-identity-sheet uses, unmodified, because the annotation problem (precisely locating panels/text in a multi-panel design sheet) is identical regardless of subject.

Always produce both. The paragraph prompt is what you feed the image model; the compositional deconstruction is what you save back to the vault so the sheet's structure survives past the single image (feeding animation, modeling, or a second illustrator's work later).

Workflow

1. Pull the character truth

Before drafting, gather (ask the user if any are missing rather than inventing them):

• Name, alias/nickname, role/function, archetype — the identity block that anchors the whole sheet (e.g. "Ding-Bot / Chime-E / Communication AI Assistant / Cheerful Guide").
• Personality & behavior — 2-4 traits plus one "core theme" line and one behavior note. This shapes expression, pose energy, and how "busy" the detail panels should feel.
• Silhouette & body type — the character's base shape (bell-shaped, humanoid, quadruped, boxy robot) since this is what has to read identically across all four turnaround views.
• Wardrobe / accessories — anything worn or carried (headphones, a hat, a tool, a cape) — these need their own detail panel if they're a defining feature.
• Color palette — primary, secondary, and accent colors, stated with role, not just a flat list (e.g. "primary: orange bell body; secondary: light-blue headphones and internal details; accent: gold/brass base metal, dark blue, white").
• Material/render style — 3D-rendered toy/vinyl-figure look, flat vector illustration, painterly concept-art, anime cel-shaded, etc. This has to be named explicitly and held constant across every panel, or Ideogram will drift style between the turnaround and the detail sheet.

If there's no existing character brief for this project, ask for the equivalent before drafting — a sheet built on invented traits is decoration, not a locked design, and will need to be redone the moment a real character brief exists.

2. Choose the panel set

Not every sheet needs every panel from references/sheet-anatomy.md — pick the ones that earn their place for this character. The canonical set, in rough reading order:

• Header/overview block — small text panels naming the character, alias, role, archetype, personality/behavior, wardrobe, and color palette (with swatch bars). This is the "spec sheet" framing that keeps the whole board from reading as fan art.
• Full body turnaround — front, 3/4, side, and back views of the character in the same neutral pose, each labeled, ideally with a height-scale ruler running alongside (inches and/or cm) to lock proportions.
• Head & detail sheet — a close-up front face, a close-up of any signature head-worn/hair detail (headphones, hat, ears), and a close-up of a signature hand/gesture or held prop.
• Accessories & interior/technical diagram — only if the character has a mechanical, robotic, or gadget-driven identity: an exploded/cutaway diagram of internal components or accessories with labeled callout lines. Skip entirely for organic/human characters unless they carry a notable prop worth diagramming.
• Color palette bar — primary/secondary/accent swatches, usually docked near the header block, each labeled.

Naming the panels explicitly in the prompt (not just "character sheet with some views") is what keeps Ideogram from defaulting to a single illustrated portrait or a loose sketch-dump instead of the structured grid you want.

3. Write the paragraph prompt

Structure, in order:

1. Frame it as a character model sheet / character turnaround sheet / character reference sheet for [name], not "an illustration of" or "art of" — the framing word is what nudges Ideogram toward the structured multi-view grid instead of a single hero pose.
2. Name the exact panel set from step 2, in the order they should read (turnaround first, then head/detail, then palette/overview, then technical diagram if included).
3. Describe the character itself once, in full (silhouette, proportions, face, wardrobe/accessories, colors with hex or named-role if known) — this single description is what has to hold consistent across every panel, so front-load it rather than re-describing the character differently per panel.
4. State the render/material style explicitly (e.g. "3D-rendered vinyl-toy style with soft studio lighting" or "flat cel-shaded vector illustration") and repeat that it must be held identical across every panel — this is the single most common failure mode (style drifting between the turnaround and the detail close-ups).
5. Color: exact hex/named values stated with role (primary body, secondary wardrobe/accent, trim/metal color) — never as an unweighted list.
6. Background/composition note: clean white or light-gray background with thin grid lines and thin black panel borders, each panel labeled with a small caption (FRONT / 3-4 VIEW / SIDE / BACK / FRONT FACE / etc.) — this is what keeps the sheet reading as a reference document, not a moodboard or collage.
7. Negative space / anti-slop: see references/anti-slop-discipline.md — no inconsistent proportions/colors between panels, no panels that drift off-model, no generic stock-mascot clichés unless the brief specifically calls for them, no invented text/labels beyond what's specified.

Keep it one paragraph (it can run longer than a logo prompt — a character sheet is carrying more distinct panels — but stay a single continuous paragraph rather than a bulleted prompt). If you want to explore genuinely different design directions (not just palette swaps), write separate prompts per direction — same "structural variety over palette-swaps" rule as logo-prompting and brand-identity-sheet.

3.5. Run the anti-slop / on-model gate before generating

Before calling generate_image, score the drafted prompt against the pre-generation gate table in references/anti-slop-discipline.md (character-specificity, style-lock transcription, off-model risk sweep, panel-labeling completeness, restraint). Anything scoring under 3 on any axis means revise the prompt now — regenerating a shipped image is expensive; revising a paragraph before you generate is not. On-model consistency (step 4 above) is the character-sheet-specific failure mode that brand-identity-sheet doesn't have to worry about as acutely, since a character's silhouette repeating correctly across four different angles is harder for an image model to hold than a wordmark repeating across two scales.

4. Generate it — don't just hand back a prompt

Call mcp__ideogram__generate_image with the paragraph prompt. Character model sheets benefit from style_type: "DESIGN" or "GENERAL" (test both if the character has a painterly/illustrative brief rather than a clean product-design one) and a landscape-to-square aspect ratio ("4x3", "3x2", or "1x1") since the canonical layout (turnaround block + head/detail block side by side) reads wider than tall. Use rendering_speed: "QUALITY" — this is a composition-heavy image with many small text-bearing panels and a proportion-lock requirement across views, and QUALITY holds up both legibility and consistency better than FAST/FLASH at that density.

If the user supplies a reference image they like (their own upload, or a prior generation) and wants variations or a new pose/detail panel added, use mcp__ideogram__remix_image against it instead of a fresh generate_image call — pass image_weight around 60-80 for "keep this exact character, adjust the sheet layout," lower for looser reinterpretations.

5. Reverse-engineer into a compositional deconstruction

Once the sheet is generated (or if the user hands you an existing reference image to work from), produce the structured JSON breakdown described in references/composition-spec-format.md. Two ways to get there:

• From a generated image: you already know the exact panel layout you asked for in step 3 — write the JSON directly from that, since you control the composition.
• From a reference image the user supplies: call mcp__ideogram__describe_image first to get the model's own read of the image, then structure that description into the high_level_description + elements schema — add bbox estimates based on visual inspection of the image, not fabricated precision. If you can't confidently place a bbox, use a coarse quadrant estimate and say so rather than inventing exact pixels. See examples/ding-bot-character-sheet.md for a full worked example at this level of detail (a real reverse-engineered spec, not a generated one, showing the target granularity: every view, every label, every scale marking captured as its own element).

Save both artifacts (prompt + JSON) to the project's character/branding folder — see step 6.

6. Save what you made

Write the paragraph prompt, the JSON compositional deconstruction, and the resulting image URL(s) to 04-projects/<project>/logo-explorations/character-sheet-<label>.md (or wherever that project keeps its branding/character work — check for an existing logo-explorations/, branding/, or characters/ folder first and match it). Follow the "No Context Lost" rule: nothing generated in this skill should live only in the conversation.

If a panel or a technique turns out to be a pattern worth reusing across characters (not just this one character's specific execution of it), add it to references/sheet-anatomy.md as a named panel type, the same way brand-identity-sheet and logo-prompting grow their reference files from real jobs.

Reference files

• references/composition-spec-format.md — the JSON schema for the compositional deconstruction (shared, unmodified, with brand-identity-sheet's schema — high_level_description, elements[] with type/bbox/desc, text elements with text).
• references/sheet-anatomy.md — the named panel types (header/overview block, full body turnaround, head & detail sheet, accessories/interior diagram, color palette bar, height-scale ruler) with what each one is for and when to include or drop it.
• references/anti-slop-discipline.md — the character-sheet-specific failure modes (off-model drift between panels, style drift between the turnaround and detail close-ups, generic stock-mascot clichés, unlabeled/mislabeled panels) and the pre-generation gate to run before every generate_image call.
• examples/ding-bot-character-sheet.md — a full worked compositional-deconstruction example (the "Ding-Bot" AI-assistant mascot sheet) at the target level of detail, showing every turnaround view, detail panel, palette swatch, and technical-diagram callout captured as its own element.