How to give your AI design taste怎么给你的 AI 设计品味
AI has no design taste of its own. What you can do is supply taste from the outside — and some ways of doing that work far better than others.AI 没有自己的设计品味。你能做的是从外部给它供给品味——而其中有些方式,远比另一些有效。
What 'taste' means for an AI对 AI 来说,“品味”是什么
Taste, in design, is a consistent set of intentional choices — a point of view applied the same way across every screen. An AI has none of this by default. Left alone it returns the average of its training data, which reads as competent but anonymous.在设计里,品味是一组一致、有意的选择——一个观点,在每一个屏幕上以同样的方式被贯彻。AI 默认完全没有这些。放任不管,它返回训练数据的平均值,读起来能用但匿名。
So 'giving your AI taste' is not about unlocking something inside the model. It is about feeding it a specific point of view from outside, in a form it can apply consistently. The question is only which form works best.所以“给你的 AI 品味”,不是去解锁模型内部的某种东西。而是从外部喂给它一个具体的观点,以一种它能一致套用的形式。问题只在于:哪种形式最有效。
Lever 1 — better prompts (weakest)杠杆一 —— 更好的 prompt(最弱)
The first lever is the prompt itself: be specific, name constraints, show one example. This is real but limited. The instruction shapes a single output and then evaporates; across many screens the model drifts back to its default, so consistency is hard to hold.第一个杠杆是 prompt 本身:写具体、点约束、给一个例子。这有用,但有限。指令塑造单次产出,然后蒸发;跨多个屏幕,模型漂回默认值,一致性很难维持。
Prompts are the right tool for a one-off and the wrong tool for a system. If you need ten slides or twenty components to feel like one product, the prompt is too thin a thread to carry the taste.prompt 适合一次性的活,不适合一个系统。如果你要十张幻灯片、二十个组件读起来像同一个产品,prompt 就是太细的一根线,扛不住品味。
Lever 2 — reference images and examples杠杆二 —— 参考图与示例
The second lever is showing, not telling: paste a screenshot, link a site you like, give a few examples to imitate. This lands better than adjectives because it points at something concrete.第二个杠杆是“给它看”,而不是“跟它说”:贴一张截图、链一个你喜欢的站、给几个可模仿的例子。这比形容词落地,因为它指向了具体的东西。
But references are lossy. The model infers what it thinks matters from a picture, and different runs infer differently — it might copy the color but miss the spacing logic, or borrow the type but not the grid. You get closer to the look without ever pinning it down.但参考是有损的。模型从一张图里推断它“以为重要”的东西,不同次推断不同——它可能抄了颜色却漏了间距逻辑,或借了字体却没要网格。你更接近那个长相,却始终没把它钉死。
Lever 3 — a machine-readable design spec (strongest)杠杆三 —— 机器可读的设计规范(最强)
The strongest lever is to hand the AI an actual design system in a form it can read: the real colors, type scale, spacing rhythm, shape language, and component rules, written down. Now the AI is not guessing at taste from a picture or an adjective — it is applying a defined system, value by value, the same way every time.最强的杠杆,是把一套真实的设计系统以模型能读的形式交给 AI:真实的配色、字阶、间距节奏、形状语言、组件规则,白纸黑字写下来。这时 AI 不再从一张图或一个形容词里去猜品味——它在套用一套已定义的系统,一个值一个值地、每次都一样地执行。
This is what closes the consistency gap. A spec does not evaporate between prompts and it does not have to be re-inferred from an image; it is the same input the model can apply to slide one and slide forty. The clearest form of this is a DESIGN.md — a single machine-readable file that carries a complete style, ready for an AI to consume.这才是补上一致性缺口的东西。规范不会在 prompt 之间蒸发,也不必从图里反复重新推断;它是同一份输入,模型既能套到第一张幻灯片,也能套到第四十张。这种形式最清晰的载体,就是 DESIGN.md——一个机器可读的单一文件,装着一整套风格,随时可被 AI 消费。
Curio exists to supply exactly this lever: each design style packaged as a spec your AI can apply directly, so you are not relying on prompt-luck for a consistent, intentional look.Curio 的存在,正是为了供给这个杠杆:每一种设计风格都打包成你的 AI 能直接套用的规范,于是你不必靠“prompt 运气”去博一个一致、有意的长相。
Putting it together把三个杠杆合起来
Use the levers in order of strength. Start from a real design spec as the backbone, lean on references where the spec is silent, and reserve prompts for the small, local adjustments. The spec carries the taste; the prompt carries the task.按强弱顺序用这些杠杆。以一份真实的设计规范作主干,在规范没说到的地方靠参考补,把 prompt 留给小而局部的微调。规范扛品味,prompt 扛任务。
The practical move: pick one documented style, give your AI its spec, and keep every screen on that same system. Taste, for an AI, is just specificity supplied consistently from outside.实操就一句:挑一个有据可查的风格,把它的规范给你的 AI,让每一个屏幕都待在同一套系统上。对 AI 而言,品味,不过是从外部一致供给的“具体”。
FAQ常见问题
Can't I just tell the AI to 'use good design'?我不能直接让 AI“用好设计”吗?
No — 'good design' is an adjective, and to a model an adjective resolves to its average. Every style in the training data claims to be good, clean, and modern, so the instruction points nowhere specific and you get the mean. Taste has to be supplied as concrete, consistent rules.不行——“好设计”是个形容词,而形容词对模型来说会解析成它的平均值。训练数据里每种风格都自称好、干净、现代,所以这条指令哪儿也不指,你拿到的就是均值。品味必须以具体、一致的规则供给。
Reference images or a design spec — which is better?参考图和设计规范,哪个更好?
A spec, for anything beyond a one-off. References are lossy and re-inferred differently each run, so consistency suffers. A spec is the same deterministic input every time, complete and reusable across many screens — which is exactly what a consistent look requires.除一次性外,都选规范。参考是有损的,每次还会被不同地重新推断,一致性受损。规范是每次都一样的确定性输入,完整且可在多个屏幕上复用——而这恰恰是“一致的长相”所需要的。
What format does the AI actually consume?AI 实际消费的是什么格式?
A machine-readable spec it can parse and apply — for Curio styles, a DESIGN.md: one file holding the full palette, type, spacing, and component logic. You hand it over by download, share link, or an MCP connection, and the AI applies the values directly.一份它能解析并套用的机器可读规范——对 Curio 风格而言就是 DESIGN.md:一个文件装下完整的配色、字体、间距与组件逻辑。你通过下载、分享链接或 MCP 连接把它交过去,AI 直接套用其中的值。