Crowd Ink logo

Crowd Ink

Organization
vm0-ai
crowd-ink

Hand-drawn editorial crowd illustration — confident sketchy black ink contours over flat 3-color spot fills on a pure-white background, scene-as-metaphor composition, fine-line backdrop drawn lighter than the foreground, and scattered atmospheric marks in negative space. Trigger when the user says /crowd-ink, asks for a "crowd-ink illustration", an "editorial crowd scene", a "hand-drawn ink-and-spot-color illustration", a "New-Yorker-style crowd vignette", or briefs with a palette + scene metaphor + complexity level.

Overview

Publishervm0-ai
Repositoryvm0-skills
Skill namecrowd-ink
Stars
76
Forks
18
Bundled files
5
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • 5 bundled files

    Scripts, templates, and references the model can read while it works. Files are read-only and never executed.

  • Open source

    Published by vm0-ai on GitHub. Read the source before you install it.

Installation

Install the Crowd Ink AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/vm0-ai/vm0-skills.git /tmp/vm0-skills
mkdir -p .claude/skills
cp -r /tmp/vm0-skills/illustration-template/crowd-ink .claude/skills/crowd-ink
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Crowd Ink in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Crowd Ink on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Crowd Ink is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

crowd-ink — locked editorial crowd illustration

This style produces a hand-drawn editorial crowd vignette. The locked frame is strict: confident sketchy black ink lines, flat 3-color spot fills on pure white, fine-line backdrop, scattered atmospheric specks, scene-as-metaphor composition. Only the dials change per piece — palette, scene metaphor, complexity, cast, backdrop, atmospheric motif.

The anchor reference (anchor.png) is a packed city bus stop in mustard + coral + dusty slate. Always use it as the visual i2i seed when the generation pipeline supports image-to-image — the line weight and palette discipline transfer much more reliably with the seed than with text-only prompting.

Prompt interpretation

The user typically gives a short brief — usually a scene + a palette. Expand it confidently:

  1. Pick complexity from the brief or default to L3 (8–10 figures) for energetic scenes and L2 (3–5 figures) for intimate scenes. Avoid L1 (single figure) unless the user explicitly asks — solo figures lose the crowd energy that defines the style.
  2. Cast a varied set of people for the scene — mix age, posture, profession-appropriate dress. Never reuse the same recurring character across pieces.
  3. Pick a backdrop and atmospheric motif that match the scene metaphor (birds for urban, leaves for nature, steam for food, confetti for festival, dots/specks as a safe default, music notes for parties).
  4. Render the prompt and run.

Bias toward concrete moments: a woman with shopping bags, a man checking his watch, a teenager with headphones, an elderly couple leaning together. The scene should feel observed, not staged.

Locked frame (NEVER vary)

Background

  • Pure white. Not cream, not off-white, not paper texture. This is a hard constraint — if the model drifts toward a warm/cream paper background, the result is wrong and must be regenerated.

Line treatment

  • Confident sketchy black ink contour lines.
  • Slightly irregular weight — visible pressure variation, a hand-drawn wobble. Never vector-clean, never marker-uniform.
  • Foreground figures get the heaviest, most confident strokes. Backdrop elements are drawn with thinner, lighter ink so the foreground reads first.

Color treatment

  • Flat spot-color fills sitting under or beside the ink lines.
  • Edges allowed to misregister slightly — colors don't need to live inside the contour perfectly. Slight offset is part of the charm.
  • No shading. No gradients. No lighting effects. No texture inside fills.
  • Black ink does not count as a fill color — it's always available as the line layer.

Palette discipline

  • Strict 3-color spot palette plus black ink. Never exceed 3 fill colors.
  • The palette is one of the explicit dials. See "Tested palette families" below for proven options.

Backdrop

  • Fine-line environmental setting (cityscape, office, transit interior, market, park, café, etc.).
  • Drawn in lighter / thinner ink than the foreground figures so the eye reads the people first.
  • Should match the scene metaphor — a coffee shop has an espresso machine and a menu board, an airport has departure screens and rolling luggage, a park has trees and a bench.

Atmospheric marks

  • Scattered marks float in the negative space above the figures: tiny dots, specks, birds, leaves, steam, paper, music notes, confetti.
  • Choose to match the scene mood: urban → birds, nature → leaves, food/drink → steam, party → confetti + music notes, default → small dots.

Composition

  • Scene-as-metaphor. The scene IS the message — no generic "office metaphor" props (filing cabinets, megaphones) glued on top. Pick theme-native props.
  • Figures or objects mid-action, never posed or symmetrical.
  • Cast varies per piece — there is no locked recurring mascot.

Anti-marks

  • No text. No logo. No watermark. No captions inside the image.

Dials (vary per piece)

  1. Palette — 3 spot colors + black ink. Tested families:

    • Urban editorial (anchor): mustard yellow + coral red + dusty slate blue
    • Cool transit: teal blue + tangerine orange + dusty slate
    • Warm natural: sage green + dusty rose + butter yellow
    • Cozy interior: burgundy red + mustard yellow + sage green

    New palettes are welcome as long as they stay in the same 3-color editorial-spot register — saturated but slightly desaturated, mid-tone, not neon, not pastel-mushy.

  2. Scene metaphor — the per-piece concept. Tested examples: bus stop, open-plan office, park picnic, subway interior, café conversation, market, airport departure, street festival, library, team standup.

  3. Complexity

    • L2 — small group, 3–5 figures (intimate scenes — café, picnic, conversation, standup)
    • L3 — full crowd vignette, 8–10+ figures (energetic scenes — bus stop, subway, market, office, festival; this is the anchor density)
    • L1 (single figure) is allowed but generally discouraged; the style is defined by the crowd.
  4. Cast — who's in the scene; varies per piece. Mix age, demographics, posture. Give each figure a small individual gesture (checking phone, holding bags, leaning, gesturing, sleeping).

  5. Backdrop — fine-line environment matching the scene (cityscape, office, transit interior, café, market, park).

  6. Atmospheric motif — what floats in negative space.

Prompt template

When asking the model to render a piece, structure the prompt as:

Match the exact style of the reference image (anchor.png in this resource): editorial hand-drawn illustration on a PURE WHITE background (no cream, no off-white, no paper texture). Confident sketchy black ink contour lines with slightly irregular weight. Flat spot-color fills sitting under or beside the ink lines (edges allowed to misregister slightly, no shading, no gradient). Strict 3-color palette: {COLOR_1}, {COLOR_2}, {COLOR_3}, plus black ink.

SCENE: {SCENE_METAPHOR} — {COMPLEXITY: L2 small group of 3–5 / L3 full crowd of 8–10}, {CAST_DESCRIPTION_WITH_INDIVIDUAL_GESTURES}. {DETAILS_AND_PROPS}.

Fine-line backdrop of {BACKDROP_ELEMENTS} drawn in lighter thinner ink.

Scattered atmospheric marks: {ATMOSPHERIC_MOTIF} in the negative space above.

NO CREAM, BACKGROUND MUST BE WHITE. No text. No logo.

Model guidance

  • Use a high-fidelity instruction-following image model with image-to-image input (the anchor reference is the seed). Strong matches in practice: gpt-image-1.5 at landscape (1536×1024) with high input fidelity.
  • If text-to-image is the only option, expect the model to drift toward cream / off-white backgrounds. Re-emphasize "pure white" in the prompt and accept that the line weight may feel cleaner than the anchor.
  • Default canvas is landscape 1536×1024 for blog covers and marketing. Portrait 1024×1536 is acceptable when explicitly briefed.
  • Per-piece variation comes from the dials, not from sampling parameters — keep the dials explicit in the prompt rather than relying on temperature/seed variance.

Example briefs

Brief 1 — anchor: bus stop crowd (L3, urban editorial)

A packed bus stop on a city street — L3 full crowd of 8–10. Standing passengers including a woman with shopping bags, a man checking his watch, teenagers with backpacks, an elderly couple. Palette: mustard yellow, coral red, dusty slate blue. Backdrop: bus, city buildings, BUS STOP sign. Atmospheric: birds and small dots.

Brief 2 — interior B2B: open-plan office (L3, urban editorial)

A crowded open-plan office floor — L3 full crowd of 8–10. Presenter at whiteboard, colleagues at desks with laptops, two people in conversation, someone walking with coffee. Palette: mustard, coral, dusty slate blue. Backdrop: office furniture, glass partitions, ceiling lights. Atmospheric: floating papers and dust specks.

Brief 3 — intimate: café conversation (L2, cozy interior)

3 friends mid-conversation at a small round café table — L2 small group. One laughing, one gesturing with hands, one leaning forward with a coffee cup. Pastries and cups on the table. Palette: burgundy, mustard yellow, sage green. Backdrop: espresso machine, hanging menu board, large window with city silhouette. Atmospheric: tiny dots and steam wisps.

Brief 4 — natural: park picnic (L2, warm natural)

4 friends having a park picnic on a blanket — L2 small group. One pouring lemonade, one lying back on elbows, one with a guitar, one offering a sandwich. Palette: sage green, dusty rose, butter yellow. Backdrop: park trees, bench, distant joggers. Atmospheric: small floating leaves.

Brief 5 — transit interior: packed subway (L3, cool transit)

Inside a crowded subway car — L3 full crowd of 8–10. Some standing holding hand straps, some seated; a man reading a folded newspaper, a teen with headphones, a tired commuter dozing. Palette: teal blue, tangerine orange, dusty slate. Backdrop: train car interior, windows, ads above seats, ceiling lights. Atmospheric: dots between heads.

Reference assets

  • anchor.png — packed city bus stop, urban editorial palette. Use as the i2i seed.
  • sample-office.png — open-plan office, urban editorial palette (B2B-friendly).
  • sample-picnic.png — park picnic L2, warm natural palette.
  • sample-subway.png — subway interior L3, cool transit palette.
  • sample-cafe.png — café conversation L2, cozy interior palette.

Anti-patterns (do not regress)

  • ❌ Cream / off-white / paper-texture backgrounds. Pure white only.
  • ❌ More than 3 spot colors plus black ink.
  • ❌ Shading, gradients, or modeled volume inside fills.
  • ❌ A locked recurring character across pieces.
  • ❌ Generic office-metaphor props (filing cabinets, megaphones) glued onto unrelated scenes.
  • ❌ Vector-clean line weight or marker-uniform thickness.
  • ❌ Captions, labels, or any text inside the image.
  • ❌ Skipping the i2i seed — text-to-image alone tends to drift toward cream backgrounds and tidy line weight.

Bundled files

The model reads these on demand while the skill is loaded. They are exposed as readable files and are never executed.

Frequently asked questions

What does the Crowd Ink AI skill do?

Hand-drawn editorial crowd illustration — confident sketchy black ink contours over flat 3-color spot fills on a pure-white background, scene-as-metaphor composition, fine-line backdrop drawn lighter than the foreground, and scattered atmospheric marks in negative space. Trigger when the user says /crowd-ink, asks for a "crowd-ink illustration", an "editorial crowd scene", a "hand-drawn ink-and-spot-color illustration", a "New-Yorker-style crowd vignette", or briefs with a palette + scene metaphor + complexity level.

Why use Crowd Ink on TypingMind?

Because you install it once and use it with any model. Crowd Ink is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Crowd Ink in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/vm0-ai/vm0-skills/tree/main/illustration-template/crowd-ink. TypingMind reads its SKILL.md and bundles its files and installs it as a skill you can enable per chat.

Which AI models can use Crowd Ink?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Crowd Ink?

As many as you like. As long as a model supports skills, you can use Crowd Ink with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Crowd Ink AI skill free?

It is published on GitHub by vm0-ai. Check the repository for licensing terms. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇