01. 現實檢查
控制優於保真度
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
遵循科學
PinePaper 的方法源於關於大型語言模型在向量圖形方面能力的新興學術研究。這些論文證明,基於文字的向量格式為語言模型提供了自然的介面。
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. 使用案例
從「場景」到「信號」
PinePaper isn't trying to replace Hollywood VFX tools. We optimize for the high-utility layer that LLMs can master today: Functional Micro-Interactions.
即時狀態信號
「錄製中」指示器、「直播」徽章或「處理中」旋轉器。簡單的程式碼定義動畫,無需影片檔案開銷即可傳達系統狀態。
PinePaper.create('circle', {animate: 'pulse'})
資料重點
圖表的上下文感知疊加層。使用代理分析資料並以程式化方式準確放置動畫箭頭、螢光筆或標註。
PinePaper.annotate(chart, {highlight: 'Q3'})
UI 原型設計
快速生成互動式 UI 模型。由於輸出基於 DOM,這些元件通常可以直接移植到生產程式碼。
PinePaper.exportSVG() → React component
04. 架構
「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).
向量空間的代理理由
經紀人會解釋你的提示, 關於位置, 形狀和動畫的原因, 然後用密碼來表示這一點.
執行層次
您執行代碼( 貼入控制台或通过 MCP 工具) 以將向量圖像移到本地 。 這讓工具保持了自由 安全 和客戶端.
斜體匯出( E)
完善自然語言反馈,然后导出為MP4,GIF,動畫SVG,或PNG.
帶上你自己的裝置 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. 基礎
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. 核心原則
為開放網路而建
用戶端核心
無需安裝。完全在瀏覽器中運作,無需伺服器往返。您的資料保留在您的裝置上——架構上的隱私。
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.
預設全球化
支援 59 種語言和所有書寫系統,包括中文、阿拉伯文、印地文和表情符號。為任何受眾創建內容。
我們正在測試一個假設
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. 常見問題
常見問題解答
什麼是 PinePaper Studio?
一個基於瀏覽器的工具,用於創建動畫 SVG 微互動。基於 Paper.js 構建,它提供了為 AI 代理和人類設計的聲明式 API,用於理解和操作向量圖形。
PinePaper Studio 是免費的嗎?
是的,它是免費軟體。完全在瀏覽器中運行,無需帳戶。所有處理都在用戶端進行——您的資料保留在您的裝置上。
AI 代理可以使用 PinePaper 嗎?
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.
什麼是 Paper.js?
一個基於 HTML5 Canvas 的成熟開源向量圖形框架。它為向量圖形提供場景圖(DOM)。PinePaper 是在其之上構建的抽象層。
支援哪些匯出格式?
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.
有哪些可用的動畫?
10 種循環動畫（Pulse、Wobble、Rotate、Bounce、Typewriter、Shake、Swing、Jelly、Fade、Scroll）加上 100+ 進入/退出預設。15 種粒子效果（sparkle、smoke、fire、rain、snow、confetti 等）。13 種頂點變形預設（fold、twist、ripple、wave 等）。時間軸關鍵影格搭配 9 種緩動函數。
準備好實驗了嗎?
無需帳戶。一切都在您的瀏覽器中運行。