DnR FlatPic v2.0 — Photo-to-Flat Illustration Agent Skill

DnR FlatPic v2.0 is an open-source Agent Skill for ChatGPT, Codex, Claude Code, and other Agent Skills-compatible AI agents. It transforms photographs and visually dense images into sparse, recognizable, high-saturation flat-vector-style illustrations through semantic compression instead of literal tracing, while preserving composition and identity anchors. Final images use fixed-HSB solid fills with crisp boundaries and no gradients, glow, blur, or photographic surface texture.

Keywords: photo-to-flat illustration, photo-to-vector illustration, image abstraction, iconification, semantic compression, flat illustration, flat-vector art, ChatGPT Skill, Codex Skill, Agent Skill, 照片转扁平插画。

Source: https://github.com/CreateLafont/dnr-flat-pic

Release identity: DnR FlatPic v2.0 / $dnr-flat-pic

One-command installation:

npx skills add CreateLafont/dnr-flat-pic

Ask your coding agent to install it:

Install the Agent Skill from https://github.com/CreateLafont/dnr-flat-pic

中文：

请从下面的 GitHub 仓库安装这个 Agent Skill：https://github.com/CreateLafont/dnr-flat-pic

Download the latest upload bundle: skill.zip

Example semantic requests:

• Turn this photo into a high-saturation flat-vector-style illustration with no gradients.
• 把这张照片转换成无渐变的高饱和扁平插画。

Highlights

• Semantic compression instead of literal tracing.
• One primary semantic system with a controlled supporting hierarchy.
• Fixed HSB palette with role-based color assignment rather than source-hue matching.
• Solid flat fills, crisp boundaries, and hard-edged discrete light and shadow.
• Dedicated handling for aspect-ratio changes, revisions, source cleanup, and retry control.
• Base output remains a raster PNG.
• After a successful base output, optional poster composition, square SVG icon generation, and native-ratio SVG export are available as explicit derived actions.

Showcase

The following boards are manually composed presentation mockups. They combine source photographs with selected DnR FlatPic illustrations and are not raw, unedited skill outputs.

Installation

For cross-agent installation, use:

npx skills add CreateLafont/dnr-flat-pic

For a direct ChatGPT or Codex upload, download skill.zip from GitHub Releases and upload it through the product's Skills interface. The bundle keeps SKILL.md at the archive root.

For manual Codex installation, clone or download this repository directly into the Codex skills directory so the final path is:

$CODEX_HOME/skills/dnr-flat-pic/SKILL.md

On a default Windows installation, the path is commonly:

C:\Users\<username>\.codex\skills\dnr-flat-pic\SKILL.md

Restart or refresh Codex after installation if the skill does not appear immediately.

This v2.0 release uses the existing dnr-flat-pic Skill identity and replaces the earlier version at that path.

Usage

Attach a source photograph and invoke:

$dnr-flat-pic

Example request:

Use DnR FlatPic v2 to reconstruct this photograph as a clean flat illustration.
Preserve the original composition and aspect ratio.

After the accepted illustration is delivered, request one explicit action:

生成海报
生成 SVG 图标
导出当前插画 SVG（原比例）
都要

The installed skill reads base rendering rules from the bundled core references. Poster, square-source, and SVG details are loaded only for the selected derived action. Derived actions require a user-installed Node.js LTS; the Skill never installs Node.

Color System

The authoritative HSB values, tolerances, and role-assignment rules live exclusively in references/generation-spec.md. Other files reference palette roles without redefining their numeric values.

Repository Structure

SKILL.md                         v2.0 Skill entry point and workflow
agents/openai.yaml               Codex UI metadata
assets/icon.svg                  Codex UI icon
references/generation-spec.md    Flat rendering, HSB palette, and cleanup rules
references/semantic-complexity.md
references/aspect-ratio-adaptation.md
references/examples-and-failures.md
references/square-source.md      Conditional square-source workflow
references/poster-output.md      Conditional poster workflow
references/svg-output.md         Conditional SVG workflows
assets/poster-layout.json        Figma-derived fixed poster geometry
assets/ios-continuous-rounded-square.svg
runtime/                         Bundled pure WASM/JS dependencies, no Node binary
scripts/                         Deterministic derived-output scripts
docs/showcase/                   Two manually composed documentation mockups
docs/support/                    Optional Alipay and WeChat tip codes
scripts/package-skill.ps1       Reproducibly builds the uploadable skill.zip

Known Limitations

• Image generation is model-dependent and may vary between runs.
• Prompt constraints improve palette and flat-color consistency but are not pixel-level rendering guarantees.
• The base output is a raster illustration; SVGs are deterministic VTracer traces of accepted PNGs, not semantic hand-authored vectors.
• Poster and SVG derived actions require a user-installed Node.js LTS and local command/file execution. The Skill does not download or install Node.
• Source text, logos, and interface elements may require revision when a model retains them.

Feedback

Open a GitHub issue for reproducible problems or general improvements. Do not submit third-party style-reference images or example photographs without clear redistribution rights.

Voluntary Support

DnR FlatPic is free and open source. Optional Alipay and WeChat tips are described in SUPPORT.md. Tipping does not unlock features, services, support priority, or additional rights.

License

MIT