详细介绍Details
Diffui 是前 Figma 工程师做的扩散式界面设计工具。官网写:用提示词一次生成多张界面方向(自称约 50 秒出 9 张),还能上传现有截图学品牌色、字体和组件,再往同风格扩页。做完可抽成设计 token,并转成不绑框架的 Web Components;也可把构建说明交给 Codex / Cursor 一类助手去实现。定价页写个人按张计费(约 17 美分一张图)、无需绑卡试用;设计转代码在 beta 期免费。Show HN(2026-07-21)约 80 分。适合想先快速比视觉方向、再交给编程助手落地的产品与设计同学。 Diffui is a diffusion-based UI design tool from a former Figma engineer. The site says a prompt can return multiple UI directions at once (it claims about nine in ~50 seconds), and you can upload screenshots so new screens follow your brand colors, type, and components. Finished work can become design tokens and framework-free Web Components, or a build brief for agents like Codex / Cursor. Pricing is per image for individuals (about 17¢), no card required to try; design-to-code is free while in beta. The 2026-07-21 Show HN landed around 80 points. Good when you want to compare visual directions fast, then let a coding agent implement.
为什么推荐Why we recommend it 收录理由:AI×设计,和库里的 Flowstep、Claude Design、Figma 助手同一类「先出视觉再交助手」,但引擎是扩散不是对话 LLM。本库没有注册 Diffui,没有出图,没有实测 MCP / Codex 插件或转代码。 Why listed: AI×design in the same family as Flowstep, Claude Design, and the Figma agent—visual first, then agent—but driven by diffusion rather than a chat LLM. This library did not sign up, generate images, or test MCP / Codex plugins or code export.
核验记录:来源、抓取与截图说明Verification log: sources, checks, and captures
主源是官网 https://diffui.ai/ ,早间 HTTP 200;多选项生成、截图学品牌、Web Components / agent 交接与按张计价照官网、定价页与 changelog 写。Show HN https://news.ycombinator.com/item?id=48995754 (2026-07-21,约 80 分)作者自述离开 Figma 做扩散式出稿。预览为 Playwright 实拍官网首屏裁 3:2 WebP。Primary source is https://diffui.ai/, HTTP 200 this morning; multi-option generation, screenshot brand learning, Web Components / agent handoff, and per-image pricing follow the site, pricing page, and changelog. Show HN https://news.ycombinator.com/item?id=48995754 (2026-07-21, ~80 points) is the founder’s own post after leaving Figma. Preview is a Playwright capture of the homepage above the fold, cropped to 3:2 WebP.