01. 现实检验
控制优先于保真度
Models like Veo, Sora, and Kling have achieved remarkable visual fidelity. They can render almost anything as pixels. But pixels are static data—you cannot easily modify a single element, interrupt a sequence mid-animation, or attach logic handlers to specific shapes after generation.
PinePaper bets on Visual Control. By exposing the vector DOM (Document Object Model), we allow AI agents to not just "dream" an image, but to structure it. This enables interruptibility, programmatic modification, and precise component reuse.
This isn't about replacing video generation—it's about filling the gap for functional micro-graphics: status indicators, data visualizations, UI components, and interactive diagrams.
The Vector Advantage
-
1
Readable Vector Space Agents can query positions, bounds, and relationships. The DOM is inspectable—enabling understanding, not just output.
-
2
Interruptibility Modify scenes mid-animation based on events. Essential for interactive dashboards and live data.
-
3
Micro-Signals Generate functional, lightweight components (spinners, badges, progress bars) rather than full-frame art.
02. 研究背景
遵循科学
PinePaper的方法基于关于大语言模型在矢量图形方面能力的新兴学术研究。这些论文表明，基于文本的矢量格式为语言模型提供了自然的接口。
Delving into LLMs' Visual Understanding via SVG
Demonstrates that LLMs can perform visual reasoning, image classification, and generation by processing SVG as text. "The textual nature of SVG provides a data modality that LLMs excel at."
VGBench: Evaluating LLMs on Vector Graphics
First comprehensive benchmark for LLM + vector graphics. Tests understanding and generation across SVG, TikZ, and Graphviz formats with 4,279 QA pairs.
LLM4SVG: Empowering LLMs for Complex Vector Graphics
Introduces semantic tokens for SVG understanding. Released SVGX dataset with 580k SVG-instruction pairs for fine-tuning language models on vector generation.
DeepSVG: Hierarchical SVG Generation & Animation
Pioneering work on neural SVG generation. Demonstrates latent space interpolation for creating smooth animations between vector keyframes.
VDLM: Visually Descriptive Language Model
Introduces "Primal Visual Description" (PVD)—an intermediate text representation between SVG and natural language that improves LLM reasoning on graphics.
StarVector: Image & Text to SVG Generation
Multimodal LLM for SVG generation from both images and text prompts. Achieves state-of-the-art on SVG-Bench for icons, logos, and diagrams.
These papers inform our design decisions but PinePaper is an independent project, not affiliated with any research institution.
03. 用例
从「场景」到「信号」
PinePaper isn't trying to replace Hollywood VFX tools. We optimize for the high-utility layer that LLMs can master today: Functional Micro-Interactions.
实时状态信号
「录制中」指示器、「直播」徽章或「处理中」旋转器。简单的代码定义动画，传达系统状态，无需视频文件开销。
PinePaper.create('circle', {animate: 'pulse'})
数据强调
图表的上下文感知叠加层。使用代理分析数据并以编程方式在重要位置精确放置动画箭头、高光或标注。
PinePaper.annotate(chart, {highlight: 'Q3'})
UI原型设计
快速生成交互式UI原型。因为输出基于DOM，这些组件通常可以直接移植到生产代码。
PinePaper.exportSVG() → React component
04. 架构
「代理交接」
Browser security sandboxes prevent external AI agents (like Claude) from directly executing code inside a running tab. We solve this with a transparent Human-in-the-Loop workflow, or via the Model Context Protocol (MCP).
关于矢量空间的代理原因
代理解释您的及时性, 理由的位置,形状,和动画, 然后表达作为代码.
执行层
您运行代码( 粘贴到控制台中或通过 MCP 工具) 来本地化矢量图形 。 这使得工具保持了自由,安全和客户端.
斜体导出( E)
完善自然语言反馈,然后导出为MP4,GIF,动画SVG,或PNG.
带上自己的设备 LLM
PinePaper also supports client-side LLMs — Chrome's built-in Gemini Nano, or any model that runs in the browser — entirely on your device, with no account and nothing sent to a server. A constrained grammar lets the model's agent talk to PinePaper's tools directly, so a prompt becomes a scene in one step. Find it under AI / Code → Assistant in the editor.
05. 基础
站在巨人的肩膀上
PinePaper is built on Paper.js—a mature vector graphics framework. Together, they provide the abstraction layer that lets agents understand and operate in vector space.
Paper.js
The Swiss Army Knife of Vector Graphics Scripting
Paper.js is a mature vector graphics scripting framework that evolved from Scriptographer—a scripting environment for Adobe Illustrator. It provides:
-
→
Vector-First Mathematics — Points, Sizes, and Rectangles as first-class citizens with operator overloading
-
→
Scene Graph / DOM — Nested layers, groups, paths, symbols—all inspectable and modifiable
-
→
Bidirectional SVG — Import, transform, animate, and export vector graphics seamlessly
-
→
Boolean Operations — Unite, intersect, subtract, divide shapes programmatically
Code that reads like geometry — agents can reason about vector space through this syntax
06. 核心原则
为开放网络而构建
客户端核心
无需安装。完全在浏览器中运行，无需服务器往返。您的数据保留在您的设备上——架构即隐私。
AI原生API
Built from the ground up to be controlled by agents. Exposes a global window.PinePaper API for programmatic generation and manipulation.
默认全球化
支持59种语言和所有书写系统，包括中文、阿拉伯语、印地语和表情符号。为任何受众创建内容。
我们正在测试一个假设
Can AI agents understand vector space and operate within it? The abstractions—PinePaper, Paper.js—are just the interface. The real question is whether agents can reason about positions, shapes, transformations, and timing to create meaningful visual output.
What we're validating:
- ? Can agents reason about spatial relationships and coordinates?
- ? Can they compose shapes, layers, and animations coherently?
- ? Does the scene graph help them self-correct visual errors?
What works today:
- ✓Relations engine — 25+ relation types (orbits, follows, springs, wiggle, event-driven state machines); behavior lives as a graph, not callbacks
- ✓AI-native — MCP server (120+ tools) + Agent Mode, plus the global
window.PinePaperAPI for programmatic control - ✓Accessible by default — WCAG 2.1 AA widget & SVG export with embedded screen-reader descriptions
- ✓100+ animation presets + keyframe timeline (easing, spatial paths, gradients) & effects (sparkle, blast, aura FX)
- ✓40+ procedural generators — flow fields, nebulae, Truchet, halftone, ribbons — seeded & remixable
- ✓Diagrams, maps & charts — flowcharts with Mermaid import/export, choropleth maps, charts from CSV/JSON
- ✓Rigging, physics & 3D — skeletons, IK, poses, bone physics, equirectangular 3D projection
- ✓Math & simulation — ODE solvers, FFT, function/parametric plotting, live dynamic systems
- ✓Export everywhere — WebM/MP4/GIF, print PDF/PNG (300/600 DPI), animated SVG (SMIL), Lottie, glTF, sprite sheets
Help us validate this — try the editor and share your experience via our feedback board.
07. 常见问题
常见问题解答
什么是PinePaper Studio？
一款基于浏览器的工具，用于创建动画SVG微交互。基于Paper.js构建，它公开了一个声明式API，旨在让AI代理和人类理解和操作矢量图形。
PinePaper Studio免费吗？
是的，这是免费软件。完全在浏览器中运行，无需账户。所有处理都在客户端进行——您的数据保留在您的设备上。
AI代理可以使用PinePaper吗？
Yes. It exposes a global window.PinePaper object for programmatic
control. We're testing whether AI can effectively understand and operate in vector space through this
interface.
什么是Paper.js？
一个成熟的HTML5 Canvas开源矢量图形框架。它为矢量图形提供场景图（DOM）。PinePaper是构建在其上的抽象层。
支持哪些导出格式？
WebM (VP9, best quality for animations), MP4 (H.264 via WebCodecs), GIF, animated SVG (SMIL), PNG up to 600 DPI, print-ready PDF (bleed + trim marks), Lottie, glTF/GLB, BVH motion capture, PNG sequence, and self-contained interactive HTML widgets. WebM for high-quality animations; MP4 for social media; PDF for print.
有哪些可用的动画？
10种循环动画（脉冲、摇摆、旋转、弹跳、打字机、抖动、摇摆、果冻、淡入淡出、滚动）加上100多种入场/退场预设。15种粒子特效（闪光、烟雾、火焰、雨、雪、彩纸等）。13种顶点变形预设（折叠、扭曲、涟漪、波浪等）。9种缓动函数的时间线关键帧。
准备好实验了吗？
无需账户。一切都在浏览器中运行。