image2_UI
Image2 UI by ONDesign — open-source UI generation toolkit for OpenAI Codex. Turn screenshots and ideas into editable frontend code.|基于 OpenAI Codex 的开源 UI
What is image2_UI?
Image2 UI by ONDesign — open-source UI generation toolkit for OpenAI Codex. Turn screenshots and ideas into editable frontend code.|基于 OpenAI Codex 的开源 UI 生成工具,将截图或想法变成可编辑、可交互的前端界面,支持断点续跑、视觉检查与图片素材复用。
Overview of image2_UI
What it is
Image2 UI is an open-source UI generation toolkit for OpenAI Codex, developed by ONDesign. It turns screenshots, design references, or text ideas into editable, interactive frontend prototypes. The project provides a command-line interface, a resumable execution runtime, image-asset management, and visual quality checks. Actual generation requires a configured Codex environment.
Key features
- Three creation workflows: Recreate (from a screenshot or reference image), Redesign (from a reference and a new product direction), Create (from a written product brief).
- Output keeps text, buttons, forms, and navigation in code; photos, product images, and illustrations remain separate assets.
- Built around OpenAI Codex: Codex executes implementation work; runs can be inspected and resumed; verification is part of the workflow with audit, comparison, and bounded fix loops.
- Multi-agent execution is available via a scheduler that coordinates specialist implementation, review, and QA work within the same run lifecycle.
- Image generation is configurable through an asset wrapper that supports the project's image-generation channels and writes asset provenance records.
- CLI commands to start, resume, inspect, validate, and compare UI work.
Use cases
- Recreate an existing screenshot or reference image by rebuilding its layout, typography, and visual details, then verifying against the original.
- Redesign a reference interface with a new product direction before implementing and checking the result.
- Create a frontend prototype from a written product brief, developing a visual direction first and then building an editable interface.
- Designers and developers can turn visual references into working frontend prototypes using OpenAI Codex.
Getting started
Requirements: Node.js 20+, Python 3.10+, an authenticated Codex CLI, a configured image-generation channel (if needed), and Playwright for browser render checks. Clone the repository and inspect available commands: bash git clone https://github.com/zhu-guli326/image2_UI.git image2-ui cd image2-ui node scripts/image2-ui --help node scripts/image2-ui doctor For a dry-run workflow plan without generating a UI: bash mkdir my-ui node scripts/image2-ui run ./my-ui --mode create --task 'A mobile coffee ordering interface with a menu and cart' --dry-run --json Once channels are configured, start a real run: bash node scripts/image2-ui run ./my-ui --mode recreate --reference ./reference.png --task 'Recreate this screen with working navigation' node scripts/image2-ui inspect ./my-ui --latest --json node scripts/image2-ui resume ./my-ui --latest
License and pricing
MIT License. Image2 UI is an independent open-source project by ONDesign, not an official OpenAI product.
Based on the project's public README and repository metadata.