Clip Agent
从长视频自动识别高光片段，裁成适合 6 个社交平台的竖屏短视频Find highlights in long videos and turn them into vertical clips for six social platforms
Overview
这是什么What It Does
一个 AI 驱动的智能视频剪辑工具，基于字幕内容分析识别黄金片段，用 YOLOv11n 人形跟踪做智能裁剪，输出适配 TikTok / Instagram / YouTube Shorts / Threads / 小红书 / Twitter 的竖屏视频。An AI video clipping tool that analyzes transcripts, identifies strong moments, tracks people with YOLOv11n, and exports vertical videos for TikTok, Instagram, YouTube Shorts, Threads, Xiaohongshu, and Twitter.
面向 内容创作者、短视频运营、播客剪辑、社交媒体团队。Built for content creators, short-video operators, podcast editors, and social media teams.
更多说明More Details
Clip Agent 解决的核心问题是：长视频里有价值的内容，手动剪成短视频太慢。它把”看长视频 → 找高光 → 裁竖屏 → 适配平台”这条链路自动化，让创作者从几个小时的手动剪辑缩短到一次提交。
除了 Web 界面，还提供 Agent CLI：直接给 Claude Code、Codex 这类 agent 调用，stdout 只输出 summary.json，适合串进自动化工作流。
Clip Agent solves a simple bottleneck: valuable moments are buried inside long videos, while manually turning them into short clips takes hours. It automates the path from watching and selecting highlights to reframing and platform delivery.
A separate Agent CLI lets coding agents call the pipeline directly and receive a clean summary.json result for larger automated workflows.
Features
核心能力Core Capabilities
基于字幕内容做多模型 AI 分析，按平台偏好不同信号选出最具传播力的片段。YouTube 看知识密度，TikTok 看情绪爆点，小红书看实用经验。Multiple models analyze transcripts and rank moments using platform-specific signals such as knowledge density, emotional peaks, or practical value.
YOLOv11n + IoU 轻量跟踪算法，动态视频面积优先、静态视频位置优先。ASD 命中时直接跟说话人，短时缺失时继续视觉跟踪上一说话人。YOLOv11n and lightweight IoU tracking keep the active subject framed, with active-speaker data taking priority when available.
TikTok / Instagram / YouTube Shorts / Threads / 小红书 / Twitter 全覆盖，4K 智能下采样到 2K，自动适配时长、宽高比和分辨率。Outputs are adapted for TikTok, Instagram, YouTube Shorts, Threads, Xiaohongshu, and Twitter, including duration, aspect ratio, and resolution.
