HotClipCAUTION
免费开源的 AI 剪辑 / 直播切片工具:长视频、直播回放、播客一键切成爆款竖屏短视频,直发抖音/快手/B站/小红书/视频号——AI 找高光金句、弹幕热度进爆点判断、自动加字幕、横屏转竖屏,本地运行无水印不上传 | Free open-source Opus Clip alternative, 100% local: AI clips long videos & livestream VODs into viral 9:16 shorts for TikTok & YouTube Shorts. No credits, no watermark, no uploads. Win/macOS/L
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
长视频、直播回放一键切成爆款竖屏短视频
简体中文 | English | 官网 | 下载 | FAQ | 反馈
AI 自动找爆点(附理由) · 横屏转竖屏 9:16 · 自动加动态字幕 · 去气口剪口头禅
无水印 · 无积分制 · 不限时长 · 素材不上传 · 连注册都不用
本地转写现支持分段保存、停止与续跑;逐句稿提供跨句全文搜索、长稿虚拟列表,以及可试听、确认和撤销的局部时间校准。开启全场画面扫描后,搜索还会命中已确认的画面描述和屏显文字,并可直接跳到对应时刻。导出自动改善不同语言的字幕阅读节奏。可选 Qwen3-ASR 本地服务与模型评估命令见 语音与长稿使用指南。
搜索没有精确命中时,逐句稿可主动开启「相近台词」:查找仅差一个字的识别结果,并明确标为近似命中,试听确认后再选段。
搜到就能选段:台词和画面命中按时间统一浏览,可按来源筛选、定位命中字词、试听上下文,再带着完整句进入选段窗口。估算时间明确标注;长稿选段按需渲染,确认加入候选后支持撤销和重做。
多人对谈开启说话人识别后,逐句稿工作台可按 S1 / S2 等说话人筛选;搜索、相近台词和选段预览会同步收窄范围。
导出与取消
导出页会显示素材检查、字幕翻译、文案生成、多版准备、编码和文件整理阶段。准备阶段即可取消,停止清理结束后可以直接重试;重复启动不会覆盖正在运行的任务。进度在文件整理完成前最高显示 99%。关闭烧录字幕、只打开 SRT 导出时,也会保留逐字稿供字幕文件使用。
常规剪切、跳剪、音频成片、合集与动态字幕编码先写目标目录内的独立临时文件,成功关闭编码器后才替换目标。失败或取消的这次编码不会把已有目标文件截断;同一批此前完成的切片
0241906b1b3dOBSERVED · 2026-10-08Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add hotclip --env HOTCLIP_LLM_API_KEY=${HOTCLIP_LLM_API_KEY} --env HOTCLIP_VISION_API_KEY=${HOTCLIP_VISION_API_KEY} -- npx -y [email protected]{
"mcpServers": {
"hotclip": {
"command": "npx",
"args": [
"-y",
"[email protected]"
],
"env": {
"HOTCLIP_LLM_API_KEY": "${HOTCLIP_LLM_API_KEY}",
"HOTCLIP_VISION_API_KEY": "${HOTCLIP_VISION_API_KEY}"
}
}
}
}Exposed tools (3)
3 read · 0 write · 0 destructive.
| Tool | Risk | Description |
|---|---|---|
clip_video | read | 全托管切片:把一个本地长视频(播客/直播回放/课程)转写、AI 找爆点并导出为可发布的竖屏短视频(烧字幕/跳剪/表情信号),返回输出目录与每条切片的标题/评分/时间码。未提供 subtitlePath 时,首次运行会自动下载本地 ASR 模型。需要环境变量 HOTCLIP_LLM_BASE_URL/HOTCLIP_LLM_MODEL(以及非本地端点的 HOTCLIP_LLM_API_KEY)。 |
detect_highlights | read | 只找爆点不出片:转写并用 LLM 挑出爆点候选,返回每条的时间码/标题/钩子/评分/理由/AI复评意见 JSON。适合先审后剪的流程。需要 HOTCLIP_LLM_* 环境变量。 |
transcribe_video | read | 字幕接入或本地转写:提供 subtitlePath 时导入已有字幕并估算字词时间;否则用端侧 ASR(SenseVoice,首次自动下载)把视频/音频转成带时间戳的逐句稿。结果有缓存,同一文件二次调用秒回。不需要 LLM 配置。 |
Trust audit
CAUTIONgrade B · trust 84/100 Install with care. The audit found things worth knowing before you trust its output.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | FAIL |
| L1 | Static analysis of the code | WARN |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- none-observed
- Network
- declared (2 observation(s))
- Shell
- none-observed
- Dependencies
- not all pinned
- Secrets in source
- none-found
Findings (20)
payload.ts
--asr-url http://127.0.0.1:8766 可选 Qwen3 本地服务
fixture-model.tar.bz2
.nojekyll
import type { TranscriptWord } from "../../shared/api-types";import type { AutomationTask } from "../../shared/api-types";import { adjustClipBoundary, adjustCandidateBoundary } from "../../shared/boundary";import type { Transcript } from "../../shared/api-types";import type { TranscriptWord } from "../../shared/api-types";默认引擎仍为 SenseVoice。Qwen3-ASR 0.6B / 1.7B 是用户管理的可选服务;HotClip 不自动安装 Python、不自动启动服务,也不将素材发送到远程地址。模型首次由服务加载时下载并缓存在本机。协议仅接受 `http://127.0.0.1:<端口>` 或 `http://[::1]:<端口>`,拒绝重定向。
启动后,在“转写引擎”中选择 Qwen3-ASR,填写 `http://127.0.0.1:8766` 并“检查连接”。界面会显示实际加载的模型、设备和对齐器状态。`--model 1.7B` 选择更大的模型;`--port` 可改端口。省略 `--aligner` 可减少模型加载,转写字词时间将明确标为估算,编辑阶段的 Qwen 校准不可用。停止客户端任务会中止等待;服务可能仍在完成当前推理,忙
pnpm cli transcribe recording.mp4 --engine qwen3 --asr-url http://127.0.0.1:8766 --json
expect(parseCliArgs(["transcribe", "source.mp4", "--engine", "qwen3", "--asr-url", "http://127.0.0.1:8766", "--restart-transcription"])).toMatchObject({ engineId: "qwen3", localServiceUrl: "http://127pcm = np.frombuffer(base64.b64decode(body["pcm"], validate=True), dtype="<f4").copy()
expect(csv.startsWith("")).toBe(true);@ffprobe-installer/ffprobe, @tailwindcss/vite, ffmpeg-static, onnxruntime-node, react, react-dom, react-icons, sherpa-onnx-node
**[v0.15.1](https://github.com/xixihhhh/hotclip/releases/tag/v0.15.1)** (2026-08-24) "Work survives, publishing teaches": **hardware-accelerated export** (VideoToolbox / NVENC / QSV with transparent x
- **[v0.12.0](https://github.com/xixihhhh/hotclip/releases/tag/v0.12.0)** (2026-08-05) "Pick sharper, open louder": **flash-forward cold opens** (flash the most explosive 0.3-1s, then cut back; auto-c
**Model requests** have deadlines covering both response headers and body: text analysis gets 3 minutes for remote services or 5 minutes for loopback services, shared across parameter fallback and emp
resources/fonts/SourceHanSansSC-Bold.otf
Gates applied: no_behavioural_pass.
0241906b1b3dfull audit observations/trust-audit/mcp-server/xixihhhh__hotclip.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-08 | 0241906b1b3d | CAUTION | B | 84 | first audit |
Questions
What is the HotClip MCP server?
免费开源的 AI 剪辑 / 直播切片工具:长视频、直播回放、播客一键切成爆款竖屏短视频,直发抖音/快手/B站/小红书/视频号——AI 找高光金句、弹幕热度进爆点判断、自动加字幕、横屏转竖屏,本地运行无水印不上传 | Free open-source Opus Clip alternative, 100% local: AI clips long videos & livestream VODs into viral 9:16 shorts for TikTok & YouTube Shorts. No credits, no watermark, no uploads. Win/macOS/L
What tools does HotClip expose?
3 in total: 3 read-only, 0 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.
Is HotClip safe to connect to an agent?
With care. The audit graded it B (84/100) and found 20 things worth knowing before you trust this server, listed below with the exact line each was found on.
What credentials does HotClip need?
It reads HOTCLIP_LLM_API_KEY and HOTCLIP_VISION_API_KEY from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How current is this page?
The grade is for one exact copy of the source (0241906b1b3d), read on 2026-10-08. The repository is watched and re-audited when it changes.