← 返回 Skills 市场

ClipMatrix — TikTok/IG AI Video Factory

Name: ClipMatrix — TikTok/IG AI Video Factory
Author: 373246784-design

作者 373246784-design · GitHub ↗ · v1.0.0 · MIT-0

cross-platform ⚠ suspicious

总下载

当前安装

版本数

在 OpenClaw 中安装

/install clipmatrix

功能描述

Real-footage batch video production for TikTok/Instagram content matrix — AI script, 5 visual styles, auto match from your library, batch run 25 accounts, au...

使用说明 (SKILL.md)

ClipMatrix — TikTok/IG 真实素材批量混剪+口播+发布矩阵

定位：一套用真实素材自动剪辑TikTok/IG短视频的生产线，适合做矩阵运营的创作者。不是模板生成——是真的从你的素材库里匹配实拍画面、AI写文案、TTS配音、自动渲染并发布。

核心能力（6个阶段）

阶段	做什么	怎么做的
M1 策略	自动分配内容方向	按方向轮换（成都→重庆→川西→北川→川南），避免相邻视频重复
M1.5 文案	AI生成TikTok口播	DeepSeek写40-60词英文钩子→正文→CTA，注入真实地点信息（非编造）
TTS 配音	文字转语音	ChatTTS原生发音，自动清洗特殊字符，支持男女声切换
M2 审核	文案质量把关	CTA完整性检查、句子长度校验、storyboard生成
M3 素材匹配	从素材库自动挑素材	场景名→中文关键词→素材文件名匹配，7天去重，短素材自动过滤
M4 渲染	HyperFrames视觉引擎	5种视觉风格可选：
		▸ `velvet` — 金色杂志封面风（城市介绍）
		▸ `soft_signal` — 暖陶土编辑式双字体（亲子/慢旅行）
		▸ `shadow_cut` — 琥珀SVG路线时间轴（路线定制）
		▸ `swiss_pulse` — 蓝色动态排版+数字滚动（种草建议）
		▸ `comparison` — 酒红VS分屏对比（旅游对比）
M5 质检	自动质量检测	黑帧（>1s自动打回）、音频电平、字幕重叠、场景数量
M6 发布	自动排期发布	Metricool API → TikTok + Instagram，随机偏移防算法检测

适用场景

✅ 旅游/美食/探店等有真实素材库的垂直领域
✅ 运营多账号矩阵（最多25个账号同时跑）
✅ 需要每天稳定产出竖屏短视频（当前产出量：日均40+条）
✅ 想要视觉风格差异化（不同账号不同风格，避免千篇一律）
⚠️ 不适合纯AI生成画面、不适合无素材库的纯文字类视频

快速开始

# 1. 安装
openclaw skills install git:373246784-design/clipmatrix

# 2. 配置
cd clipmatrix
cp config.yaml.example config.yaml
# 编辑 config.yaml 填入: DeepSeek API Key, Metricool Token, 素材库路径

# 3. 跑一条
python3 scripts/run_and_notify.py 00 2026-06-01 AM

# 4. 批量生产（7天×2条=14条）
python3 scripts/batch_runner.py 00 2026-06-01 2026-06-07

配置

配置项	说明	必需
`DEEPSEEK_API_KEY`	DeepSeek API Key（环境变量）	✅
`METRICOOL_TOKEN`	Metricool API Token	✅
`api.deepseek.model`	文案模型（默认 deepseek-v4-pro）	-
`paths.library_dir`	素材库存放路径	✅
`workflow.tta_max_duration_sec`	TTS最长秒数（默认50）	-
`video.min_words/max_words`	口播词数范围（60-120）	-
`directions`	内容方向列表	-
`accounts.id_range`	账号ID范围	✅

故障排查

M4黑屏 → 素材太短，降 workflow.storyboard_padding 或换长素材
DeepSeek超时 → API不稳定，切 fallback_model: deepseek-v4-flash
M3素材缺口 → 素材文件名需包含场景中文名，补到 library_dir/竖屏/
M6发布失败 → Metricool Token过期，重新获取

完整排查见 references/TROUBLESHOOTING.md，架构说明见 references/WORKFLOW.md。

安全使用建议

Install only if you are comfortable granting this skill access to your social publishing accounts, Metricool token, local media library, and some OpenClaw credential/profile files. Review or remove the host-level cleanup/proxy code before running it on your main machine, and run it in an isolated environment with a dedicated config, dedicated workspace directory, and limited API credentials.

能力标签

cryptorequires-sensitive-credentials

能力评估

⚠ Purpose & Capability

Core video generation, QA, Metricool scheduling, and TikTok/Instagram publishing are consistent with the stated purpose, but the package also includes Cloudflare site analytics for a hard-coded Panda-Journeys domain and Metricool dashboard creation/sync code that are not clearly disclosed in the manifest.

⚠ Instruction Scope

Runtime instructions describe batch media production and publishing, but several scripts perform broader actions such as force-killing Chrome, killing MacPacketTunnel, disabling Wi-Fi web proxy settings, reading shared OpenClaw auth profiles, and falling back to a second LLM provider.

ℹ Install Mechanism

The install path and dependencies are ordinary for a Python media automation skill, and VirusTotal and static scan telemetry were clean; however the documentation omits some sensitive runtime side effects and auxiliary integrations.

⚠ Credentials

Use of ffmpeg, Node/HyperFrames, API calls, local media files, and Metricool credentials is proportionate, but host-wide process termination, proxy mutation, broad credential-store reads, configurable API destinations, and workspace deletion are broader than needed without strong scoping.

⚠ Persistence & Privilege

The skill writes install/license state, checkpoints, schedules, QA history, output files, and cleanup markers, which is expected for a batch workflow, but cleanup can delete a configured workspace tree and credential/session access is not limited to skill-owned secrets.

如何使用

确保已安装 OpenClaw（本地或 Docker 部署）
在对话框中输入安装命令：/install clipmatrix
安装完成后，直接呼叫该 Skill 的名称或使用 /clipmatrix 触发
根据 Skill 的参数说明提供必要输入，即可获得结构化输出

版本历史

v1.0.0

Initial release: 6-stage automated video production pipeline (M1-M6). AI script generation, TTS dubbing, footage matching, 5 HyperFrames visual styles, QA, and auto-publishing to TikTok/Instagram. Supports 25-account matrix operation.

元数据

Slug clipmatrix

版本 1.0.0

许可证 MIT-0

累计安装 0

当前安装数 0

历史版本数 1

常见问题