@enslo/sd-metadata-lite

🌐 日本語版はこちら

🔗 Live Demo

A lightweight, read-only metadata parser for AI-generated images. Designed for bookmarklets and userscripts where bundle size matters.

• Tiny: 6,911 bytes minified IIFE build
• Read-only: Extracts metadata and returns A1111-format text
• Zero dependencies: Works in Node.js and browsers
• 18+ tools: Supports all major AI image generation tools

[!TIP] Need write support, format conversion, or structured metadata? Use the full library instead.

Installation

npm install @enslo/sd-metadata-lite

Usage

ESM

import { parse } from '@enslo/sd-metadata-lite';

const response = await fetch(imageUrl);
const data = new Uint8Array(await response.arrayBuffer());
const text = parse(data);

if (text) {
  console.log(text);
  // masterpiece, best quality, 1girl
  // Negative prompt: lowres, bad quality
  // Steps: 20, Sampler: Euler a, CFG scale: 7, Seed: 12345, ...
}

Userscript

Load the IIFE build via @require:

// ==UserScript==
// @name        My Script
// @namespace   https://example.com
// @require     https://cdn.jsdelivr.net/npm/@enslo/[email protected]/dist/index.global.js
// ==/UserScript==

const response = await fetch(imageUrl);
const data = new Uint8Array(await response.arrayBuffer());
const text = sdml.parse(data);

if (text) {
  console.log(text);
}

[!TIP] Always pin to a specific version in @require for stability.

Bookmarklet

Most sites block external script loading via CSP. Instead, paste the IIFE bundle inline:

1. Copy the contents of dist/index.global.js (or download from npm)
2. Build your bookmarklet with the code inlined:

javascript:void(async()=>{/* paste contents of index.global.js here */;/* your code using sdml.parse() */})()

Since the IIFE assigns to var sdml, calling sdml.parse() works immediately after the inlined code.

API

parse(input: Uint8Array | ArrayBuffer): string

Parses metadata from an image and returns A1111-format text.

Parameters:

• input - Image file data (PNG, JPEG, or WebP)

Returns:

• A1111-format metadata string if metadata is found
• For PNGs with no generation parameters, the C2PA Content Credentials generator name (e.g. OpenAI Media Service API, Google C2PA Core Generator Library) when present
• Empty string ("") if no metadata is found or the format is unrecognized

Output Format

The output follows the Stable Diffusion WebUI (A1111) text format:

masterpiece, best quality, 1girl
Negative prompt: lowres, bad quality
Steps: 20, Sampler: Euler a, Schedule type: Karras, CFG scale: 7, Seed: 12345, Size: 512x768, Model hash: a1b2c3d4, Model: model.safetensors

Prompt and parameter fields are included only when present in the source metadata.

Supported Tools

Reads PNG, JPEG, and WebP metadata from:

• Stable Diffusion WebUI (A1111)
• Forge / Forge Classic / Forge Neo
• reForge / EasyReforge
• NovelAI
• ComfyUI
• SD.Next
• InvokeAI
• SwarmUI
• Stability Matrix
• Civitai (JPEG only verified; PNG/WebP are experimental *)
• TensorArt
• HuggingFace Space
• Ruined Fooocus
• Fooocus *
• Easy Diffusion *

[!NOTE] * Implemented based on reference code analysis, not verified with actual sample files.

Content Credentials (C2PA)

PNGs from commercial tools that sign their output with C2PA Content Credentials are detected by their generator name instead of generation parameters:

• OpenAI ChatGPT
• Google Gemini

For these images parse() returns the generator name (e.g. OpenAI Media Service API). Detection is best-effort: the credentials are easily stripped by re-saving or re-uploading, and a present manifest is not cryptographically verified.

Globals

When loaded via <script> tag or userscript @require, the IIFE build exposes:

window.sdml.parse(input)

Contributing

See CONTRIBUTING.md for guidelines.

License

MIT