home / skills / project-n-e-k-o / n.e.k.o / 3d-interaction

3d-interaction skill

/.agent/skills/3d-interaction

This skill helps you implement accurate 3D camera interactions in Three.js by dynamically mapping pixels to world space and enforcing visible boundaries.

npx playbooks add skill project-n-e-k-o/n.e.k.o --skill 3d-interaction

Review the files below or copy the command above to add this skill to your agents.

Files (1)
SKILL.md
3.8 KB
---
name: 3d-camera-interaction
description: Three.js 中处理 3D 模型拖拽、缩放、边界检测的正确方法。解决鼠标移动与模型移动不同步、缩放后只能看到模型一部分等问题。
---

# 3D 相机交互:拖拽与边界检测

## 症状

- 缩放后拖拽模型,鼠标移动 100px 但模型移动的屏幕距离不是 100px
- 放大模型后只能看到腿/身体的一部分,无法正常平移
- 拖动开始时模型位置"跳变"

## 根本原因

### 原因 1: 固定 panSpeed 导致移动不同步

**问题**: 使用固定的 `panSpeed = 0.01` 进行平移计算
```javascript
// ❌ 错误方式
const panSpeed = 0.01;
newPosition.add(right.multiplyScalar(deltaX * panSpeed));
```

**为什么发生**: 相机距离变化时,同样的世界空间距离在屏幕上的像素表现不同。距离近时像素多,距离远时像素少。

**解决方案**: 根据相机距离和 FOV 动态计算像素→世界空间的映射

```javascript
// ✅ 正确方式:动态计算
const cameraDistance = camera.position.distanceTo(modelCenter);
const fov = camera.fov * (Math.PI / 180);
const screenHeight = renderer.domElement.clientHeight;
const screenWidth = renderer.domElement.clientWidth;

// 在相机距离处,视口的世界空间高度
const worldHeight = 2 * Math.tan(fov / 2) * cameraDistance;
const worldWidth = worldHeight * (screenWidth / screenHeight);

// 每像素对应的世界空间距离
const pixelToWorldX = worldWidth / screenWidth;
const pixelToWorldY = worldHeight / screenHeight;

// 应用:鼠标移动的像素 × 每像素对应的世界空间距离
newPosition.add(right.multiplyScalar(deltaX * pixelToWorldX));
newPosition.add(up.multiplyScalar(-deltaY * pixelToWorldY));
```

### 原因 2: 基于中心点的边界限制

**问题**: 使用模型中心点的 NDC 坐标判断是否出界
```javascript
// ❌ 错误方式:限制中心点位置
const ndc = position.clone().project(camera);
if (ndc.y > 0.2) clampedY = 0.2; // 限制顶部
```

**为什么发生**: 模型放大后,中心点在屏幕中心,但身体大部分已超出屏幕。限制中心点 = 限制只能看到身体中间部分。

**解决方案**: 计算模型在屏幕上的可见区域(像素),只在可见区域过小时才校正

```javascript
// ✅ 正确方式:基于可见像素
const MIN_VISIBLE_PIXELS = 50;

// 1. 计算模型包围盒并投影到屏幕
const box = new THREE.Box3().setFromObject(vrm.scene);
const corners = [/* 8个顶点 */];

let modelMinX = Infinity, modelMaxX = -Infinity;
let modelMinY = Infinity, modelMaxY = -Infinity;

corners.forEach(corner => {
    const projected = corner.clone().project(camera);
    const screenX = (projected.x * 0.5 + 0.5) * screenWidth;
    const screenY = (-projected.y * 0.5 + 0.5) * screenHeight;
    // 更新边界...
});

// 2. 计算可见区域
const visibleWidth = Math.max(0, Math.min(screenWidth, modelMaxX) - Math.max(0, modelMinX));
const visibleHeight = Math.max(0, Math.min(screenHeight, modelMaxY) - Math.max(0, modelMinY));
const visiblePixels = visibleWidth * visibleHeight;

// 3. 只在可见区域太小时校正
if (visiblePixels < MIN_VISIBLE_PIXELS) {
    // 将模型拉回可见区域
}
```

## 关键公式

### 像素到世界空间转换
```
worldHeight = 2 × tan(fov/2) × cameraDistance
pixelToWorld = worldHeight / screenHeight
```

### 世界坐标到屏幕坐标
```javascript
const ndc = worldPos.clone().project(camera);
const screenX = (ndc.x * 0.5 + 0.5) * screenWidth;
const screenY = (-ndc.y * 0.5 + 0.5) * screenHeight; // Y 轴反向
```

## 关键经验

- 📐 **相机距离影响一切**: 所有像素↔世界空间的转换都需要考虑相机距离
- 🔲 **使用包围盒而非中心点**: 边界检测应基于模型实际占用的屏幕区域
- 🔄 **与 2D 保持一致**: Live2D/VRM 等不同类型模型应使用相同的交互逻辑阈值

Overview

This skill explains correct Three.js camera interaction for dragging, zooming, and boundary detection of 3D models. It fixes mismatches between mouse movement and model movement, prevents partial visibility after zoom, and eliminates position jumps when dragging begins. The guidance centers on pixel→world conversions and bounding-box based visibility checks.

How this skill works

It computes a dynamic pixel-to-world mapping using camera distance and field of view so mouse deltas translate to consistent world-space movement. It projects the model's bounding-box corners to screen space to measure actual visible pixels, then only clamps or recenters the model when the visible area drops below a threshold. Additional guards prevent initial jump by using the drag start point transformed into the same coordinate space as subsequent movement.

When to use it

  • Dragging or panning 3D models with an orbital or perspective camera
  • Zooming models where scale or camera distance changes screen-space mapping
  • Preventing parts of a model (limbs, torso) from disappearing after zoom
  • When drag starts show a sudden model jump or inconsistent movement
  • Implementing robust UI for VRM/Live2D/character models in Three.js

Best practices

  • Compute pixelToWorld using worldHeight = 2 * tan(fov/2) * cameraDistance and map by screen dimensions
  • Use the model's Box3 corners projected to NDC → screen to derive visible pixel bounds instead of relying on the center point
  • Only apply clamping or re-centering when visible pixel area falls below a small threshold (e.g., MIN_VISIBLE_PIXELS)
  • On drag start record the intersection point or model-space offset to avoid position jumps
  • Keep math in world-space and convert inputs (mouse deltas) once per frame to avoid cumulative errors

Example use cases

  • A character viewer where users can zoom and pan without losing limbs off-screen
  • An editor for placing accessories on avatars that requires pixel-accurate drag behavior
  • A metaverse scene where many models share consistent interaction thresholds
  • Replacing a broken fixed panSpeed implementation to support variable camera distance

FAQ

Why does my model jump when I start dragging?

You likely compute movement from screen deltas without aligning the drag start point to model/world coordinates. Fix by recording the initial intersection/offset in the same space you apply subsequent translations.

Can I still use a fixed panSpeed?

Fixed panSpeed causes inconsistent screen movement when camera distance or FOV changes. Use pixel→world mapping based on cameraDistance and fov for stable, predictable panning.