Skip to content
SkillScript-led4,268 tokens in use

character-refs

Generates a set of reference images for any story character from one description: a front full-body anchor, then headshot, profile, back and detail views.

6,884 stars on the repository Apache License 2.0eternityspring/shuohao-skills · skills/character-refs
git clone https://github.com/eternityspring/shuohao-skills.git cd shuohao-skills ./scripts/install.sh
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing character-refs costs

305tokens idle
4,268tokens in use
10,125tokens, up to

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Up to adds the files its instructions link. Counts are approximate, made offline with cl100k_base.

Model dependence Script-led

Script-ledMixedModel-led

Most of this skill is scripts, so it behaves much the same whichever model runs it.

This describes the skill's structure, how much of it is scripts. It does not say how well it works. Scripts: scripts/character-refs.mjs, scripts/core.mjs, scripts/i18n.mjs, scripts/looks.mjs, scripts/models.mjs, scripts/png.mjs, scripts/selftest.mjs.

Touches

shelldeclared and detectednetworkdetectedwrites filesdeclared and detectedcredentialsdetected

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Good
  • Frontmatter12/15
  • Structure22/25
  • Examples12/15
  • Output0/15
  • Size11/15
  • Resources3/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Good
  • Runs shell commands−2

    does this on the machine, as the frontmatter declares

    SKILL.md, line 73
  • Writes files−4

    does this on the machine, as the frontmatter declares

    SKILL.md, line 124
  • Reaches the network−10

    does this on the machine, and the frontmatter does not say so

    scripts/models.mjs, line 138
  • Reads credentials−15

    does this on the machine, and the frontmatter does not say so

    scripts/selftest.mjs, line 323

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Good
  • Development activity92
  • Maintenance52
  • Community60
  • Maturity20
  • Momentum40

character-refs turns a short character description into real reference images. The description is split into fields, gaps are filled and shown for your confirmation, then a front full-body anchor image is generated. Every other view, such as headshot, 90-degree profile, back and detail close-ups, is generated from that one anchor, in tiers you choose.

Each image is labeled, can be regenerated alone, and is marked stale when the anchor, style or description changes. An HTML report shows the whole set. It works with Qwen Image on ComfyUI, Codex image generation, the OpenAI image API or a custom command.

What You Get

  • An asset file holding the character description and every version of every image
  • A front full-body anchor and, by default, a headshot, a 90-degree profile and a back view
  • Optional detail images for hair, collar, sleeves, shoes or named features, and an optional 45-degree headshot
  • Images labeled with character, view, version, model and the references they used
  • An HTML report with stale images flagged, in Chinese, English or Japanese

Common Use Cases

  • Making consistent reference images for a short drama or anime character before video generation
  • Designing a character for your own original story without any source novel
  • Regenerating one bad view without redoing the whole set

What Makes It Unique

The whole set has a single root: only the anchor is generated from text, and each other view is generated from the anchor, so one bad image never cascades. Models can be mixed between anchor and derived views. Frozen items are tracked by hash, and changing the anchor, style or description marks dependants as stale and keeps all old versions.

Under The Hood

The skill runs a bundled Node script with no npm dependencies. It writes intake, asset, image and report files into the output folder. Image generation goes out to the backend you configure: your ComfyUI server (with optional credentials from environment variables or an env file), the local Codex image tool, or the OpenAI images API with your key, or a custom command you provide, so your prompts are sent to that service. Settings are saved in a local config file. The OpenAI adapter was only tested against a local fake server according to the skill.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers