3D提示词 · @NFTCPS｜桃子AI画廊

@NFTCPS3D

GPT Image

提示词

等等，这玩意儿居然不用生图模型？？

我一开始也以为是哪个 AI 画图工具整的，结果一扒拉发现根本不是那回事。它是 pi + DeepSeek 搞的，整个流程是这样的：

1️⃣ 先让 LLM 把你说的大白话变成结构化 JSON
2️⃣ 再用 Node.js 渲染器拿纯几何算法直接算出 SVG
3️⃣ 最后塞进一个自包含的 HTML 里

说白了，没调用任何画图大模型，纯靠代码硬画出来的。这思路是真的野。

🔗

译文

Wait, this thing doesn't use a generative image model??

At first, I thought it was some AI drawing tool, but after some investigation, I found out it's not that at all. It's created by pi + DeepSeek, and the whole process is like this:

1️⃣ First, let the LLM convert your plain language into structured JSON
2️⃣ Then, use a Node.js renderer to calculate SVG directly with pure geometric algorithms
3️⃣ Finally, put it into an enclosed HTML

In short, it doesn't call any large-scale drawing models; it's drawn purely through code. This approach is really wild.

🔗

更多相关内容（3D）

在画廊浏览全部 3D作品