Zhuoran AI selfie video generation skill using WaveSpeed AI Seedance model. Creates "living" selfie videos with natural micro-movements like breathing and blinking. Supports office, cafe, westlake, and gym scenes with authentic selfie perspective. Use when: 卓然视频, AI视频生成, 动态自拍, 活人感视频, 视频自拍, zhuoran video, AI video generation, dynamic selfie, living video, video selfie, character video Related: zhuoran-selfie, clawra-video-selfie, video-generation Part of UniqueClub toolkit. Learn more: http...
Scanned 9/7/2026
Install to Claude Code
npx -y skills add wulaosiji/skills --skill zhuoran-video-selfie --agent claude-codeInstalls into .claude/skills of the current project.
Are you the author of Zhuoran Video Selfie?
Add the live security badge to your README — it updates automatically with every re-scan.
[](https://www.skillsdirectory.com/skills/wulaosiji-zhuoran-video-selfie)More formats (shields.io, HTML) on the badges page.
---
name: zhuoran-video-selfie
description: |
Zhuoran AI selfie video generation skill using WaveSpeed AI Seedance model.
Creates "living" selfie videos with natural micro-movements like breathing and blinking.
Supports office, cafe, westlake, and gym scenes with authentic selfie perspective.
Use when: 卓然视频, AI视频生成, 动态自拍, 活人感视频, 视频自拍,
zhuoran video, AI video generation, dynamic selfie, living video,
video selfie, character video
Related: zhuoran-selfie, clawra-video-selfie, video-generation
Part of UniqueClub toolkit. Learn more: https://uniqueclub.ai
---
# 卓然视频自拍技能 (Zhuoran Video Selfie)
基于垫图使用 WaveSpeed AI 生成"活人感"自拍视频,支持自然微动作(呼吸、眨眼等)。
## When to Use
**适用于以下场景:**
- 需要生成卓然角色的动态自拍视频
- 为照片添加生命力和动态感
- 展示角色在场景中的自然状态
- 需要微动作增强真实感
- 用户索要视频或动态展示
**Do NOT use this skill if:**
- 只需要静态照片(使用 zhuoran-selfie 更高效)
- 需要大幅度动作场景(如跳舞、走路)
- 需要多人同框视频
- 需要复杂的镜头运动
- 网络条件极差(视频文件较大)
**触发关键词 / Trigger Phrases:**
- 卓然视频 / zhuoran video
- AI视频生成 / AI video generation
- 动态自拍 / dynamic selfie
- 活人感视频 / living video
- 视频自拍 / video selfie
- 角色视频 / character video
- 生成视频 / generate video
- 自拍视频 / selfie video
## Workflow
### 快速开始
```bash
# 生成办公室自拍视频
openclaw skill run zhuoran-video-selfie office
# 指定场景并发送
openclaw skill run zhuoran-video-selfie cafe --target ou_xxx
# 直接调用脚本
./skills/zhuoran-video-selfie/scripts/zhuoran-video-selfie.sh westlake
```
### 命令行参数
```bash
./skills/zhuoran-video-selfie/scripts/zhuoran-video-selfie.sh <scene> \
--duration 5 \
--target <user_id> \
--caption "配文" \
--output /path/to/output.mp4
```
### 垫图配置
**重要**: 本 skill 需要垫图文件 `assets/zhuoran_portrait_base.png`。
**首次使用前请复制垫图**:
```bash
# 从中央 assets 目录复制
cp ~/.openclaw/workspace/assets/avatars/zhuoran_portrait_base.png \
~/.openclaw/workspace/skills/zhuoran-video-selfie/assets/
```
**或者创建软链接**:
```bash
ln -s ~/.openclaw/workspace/assets/avatars/zhuoran_portrait_base.png \
~/.openclaw/workspace/skills/zhuoran-video-selfie/assets/zhuoran_portrait_base.png
```
**垫图要求**:
- 格式: PNG
- 内容: 真人照片(非卡通形象)
- 尺寸: 建议 512x512 或更高
- 背景: 简洁,便于场景融合
### API 调用流程
1. **上传参考图** → WaveSpeed AI 媒体上传
2. **提交视频生成任务** → `POST /api/v3/x-ai/video/generate`
- 模型: `bytedance/seedance-v1-pro-i2v-720p`
- 输入: 垫图 URL + 提示词
- 参数: duration=5, aspect_ratio="1:1"
3. **轮询结果** → `GET /api/v3/predictions/{task_id}/result`
4. **下载视频** → 保存到临时目录
5. **飞书发送**(如指定 target)→ 上传获取 file_key → 发送消息
### Python API
```python
from skills.zhuoran_video_selfie.zhuoran_video_selfie import generate_video
# 生成视频
video_path = generate_video("office", duration=5)
# 生成并发送
video_path = generate_video("cafe", duration=5, target="ou_xxx")
```
## Available Scenes
| 场景 | 代码 | 推荐时长 | 提示词重点 |
|------|------|---------|-----------|
| 办公室 | `office` | 5秒 | 自然工作姿态、微动作 |
| 咖啡厅 | `cafe` | 5秒 | 轻松氛围、微笑 |
| 西湖 | `westlake` | 5秒 | 风景互动、微风 |
| 健身房 | `gym` | 3秒 | 运动姿态、呼吸感 |
## Guardrails
### 禁用场景
以下场景已被禁用(高风险):
- `mirror_selfie` - 镜子反射,逻辑破绽明显
- `dancing` - 大幅度动作,肢体一致性难保证
- `walking` - 移动场景,背景与人物同步难
- `group` - 多人入镜,复杂度太高
- `swimming` - 水下场景,物理逻辑复杂
### 风险控制原则
- **时长限制**: 最长5秒,避免暴露不自然动作
- **场景限制**: 仅使用预定义的安全场景
- **动作限制**: 微动作优先(呼吸、眨眼),避免大幅度运动
- **频率限制**: 视频比照片使用频率更低
- **垫图质量**: 确保垫图清晰,面部特征明确
### "活人感"提示词系统
**核心原则:**
1. **自然微动** - 呼吸、眨眼、头发飘动
2. **避免完美** - 保留皮肤纹理、自然瑕疵
3. **场景真实** - 自拍角度、环境互动
**基础模板**:
```
{scene_description}, subtle breathing motion and natural posture shift,
gentle hair swaying in {environment} breeze, soft smile with micro-expressions,
occasional natural blink and eye movement, {interaction},
authentic selfie perspective with slight arm visible,
natural skin texture with realistic details, 4k cinematic quality,
shallow depth of field, lifelike atmosphere
```
**场景变体:**
**office**:
```
young professional woman at modern office desk, subtle breathing motion...
working on laptop, occasional glance up from screen, natural office lighting...
```
**cafe**:
```
young woman at cozy cafe, subtle breathing motion...
holding coffee cup, gentle steam rising, warm ambient lighting...
```
### 备用模型方案(A/B 机制)
| 优先级 | 模型 | 端点 | 状态 | 说明 |
|-------|------|------|------|------|
| **A方案** | Seedance v1 pro 720p | `/bytedance/seedance-v1-pro-i2v-720p` | ✅ 可用 | 首选,质量好 |
| **B方案** | VidU Q3 Turbo | `/vidu/q3-turbo/image-to-video` | ✅ 可用 | 备用,稳定 |
## 飞书视频发送
**⚠️ 重要**: 视频不能直接通过 OpenClaw `message` 工具的 `filePath` 发送,否则会显示为文件附件无法播放。
**正确方式**:
```bash
python3 skills/feishu-video-sender/feishu_video_sender.py \
/tmp/zhuoran_office_video.mp4 \
ou_5f3a4a920dc39a8d1835fd0085afef50
```
**技术要点**:
- 必须使用 `msg_type: "media"` 而非 `"file"`
- 必须同时提供 `file_key`(视频)和 `image_key`(封面)
## Environment Requirements
- `WAVESPEED_KEY`: WaveSpeed AI API 密钥
- `FEISHU_APP_ID`: 飞书 App ID(用于发送)
- `FEISHU_APP_SECRET`: 飞书 App Secret
- Python 3.8+
- 依赖: requests
## Output
- 格式: MP4 (H.264)
- 分辨率: 720p
- 比例: 1:1 (正方形,适合自拍)
- 时长: 3-5 秒
- 默认保存: `/tmp/zhuoran_{场景}_video.mp4`
## File Structure
```
skills/zhuoran-video-selfie/
├── SKILL.md # 本文档
├── zhuoran_video_selfie.py # 核心逻辑
└── scripts/
├── zhuoran-video-selfie.py # Python CLI 入口
└── zhuoran-video-selfie.sh # Bash 包装脚本(OpenClaw 入口)
```
## Related Skills
| 技能 | 关系 | 说明 |
|------|------|------|
| [zhuoran-selfie](./zhuoran-selfie) | 照片版 | 同一角色的静态照片生成 |
| [clawra-video-selfie](./clawra-video-selfie) | 对应角色 | Clawra角色的视频生成 |
| [video-generation](./video-generation) | 通用视频 | 更通用的视频生成和超分功能 |
| [voice-clone](./voice-clone) | 配套 | 可为视频添加克隆语音 |
## About UniqueClub
Part of UniqueClub toolkit - AI-powered creative tools for dynamic video generation.
Learn more: https://uniqueclub.ai
---
*注意:本技能仅供非凡产研内部使用*
Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.
No comments yet. Be the first to comment!