AutoClip - AI Video Clipping & Highlight Tool
AI-powered video clipping and highlight generation with automatic download from YouTube and Bilibili.
- Source repo
- zhouxiaoka/autoclip
- Stars
- ★ 9.5k
- Last updated
- 3d ago
- License
- MIT
- Primary language
- Python
- FA score
- 45/100 · Major gaps
At a glance
- How it runs
- Works with
- Portable with changes
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- Content creators who need to quickly extract highlights from long videos for social media.
- Not a fit if
- Users wanting a ready-made SaaS without deploying Redis and FFmpeg themselves
- Teams needing Bilibili upload, subtitle editing, or multi-language support, still in development
- Source review
- 45/100 · Major gaps 1 safety controls not found
What does this agent do, and when should you use it?
AutoClip is an open-source intelligent video clipping system that downloads videos from platforms like YouTube and Bilibili, analyzes content using the Qwen large language model, automatically identifies highlight segments, and generates compilations. The system uses a modern front-end/back-end separation architecture, with FastAPI and Celery for backend processing and React/TypeScript for the frontend, providing real-time progress updates and comprehensive project management. Deployment is straightforward with Docker one-click scripts and local startup scripts, offering flexibility for different production environments. Advanced features such as Bilibili upload and subtitle editing are currently under development.
AutoClip ingests video files or platform links, downloads YouTube videos via yt-dlp, downloads Bilibili videos through custom APIs, and supports uploading local files. The video is submitted to an AI analysis engine (based on Qwen) that extracts an outline, identifies topic timelines, scores segments for highlight-worthiness, and generates titles. Users can create projects in the React frontend, monitor processing progress (via WebSocket), manage generated clips and compilations, and export or download results. The backend utilizes Celery for asynchronous tasks, Redis as the message broker, SQLite for data storage, and exposes a REST API via FastAPI.
- Content creators who need to quickly extract highlights from long videos for social media.
- Video editors looking to automate batch processing to reduce manual clipping time.
- Bilibili uploaders wanting to upload sliced videos, though the upload feature is not yet available.
- Educational institutions needing to extract key segments from lecture recordings.
- Entertainment enthusiasts wanting to auto-generate highlight compilations from movies or shows.
- Business teams extracting decision points from meeting recordings for review.
How do you install or deploy this agent?
Requirements: Python 3.8+, Node.js 16+, Redis 6.0+, FFmpeg, Docker (optional). Recommended: clone the repository and run ./docker-start.sh for Docker deployment. For local setup, create a virtual environment, install dependencies, configure a .env file with Qwen API key, and then start Redis, backend, and frontend. See the 'Quick Start' section in the README for exact commands.
How do you use this agent?
After starting the services, access the frontend at http://localhost:3000. Click 'New Project', select a video source (YouTube link, Bilibili link, or local file), paste the link or upload the file (optionally with subtitles), and start the download. The system automatically performs AI analysis, clip generation, and compilation. In the project detail page, you can review clips, edit titles and descriptions, create/edit compilations, and download outputs. API documentation is available at http://localhost:8000/docs.
What are this agent's strengths and limitations?
- Modern tech stack (FastAPI, React, TypeScript) with a clean separation of concerns, making it easy to extend.
- Supports multi-platform video download (YouTube, Bilibili) and local upload, with integrated AI analysis for high automation.
- Offers Docker one-click deployment and detailed local deployment docs for various environments.
- Asynchronous task queue with real-time progress updates improves user experience.
- AI analysis depends on Alibaba Cloud's Qwen API, creating vendor lock-in and requiring a paid API key.
- Bilibili upload, subtitle editing, and multi-language support are not yet implemented, missing some advanced features.
- Backend storage relies on SQLite and Redis, which may need migration to PostgreSQL for large-scale deployments.
- Mobile support is under development; the current interface is desktop-focused.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| AutoClip - AI Video Clipping & Highlight Tool This agent | 45 · Major gaps | Self-hosted serviceFree + model costs | ★ 9.5k | 3d ago | Python | — |
| GOD Agent Society Control Room | 60 · Some gaps | CLIFree + model costs | ★ 1.2k | 16d ago | Python | OpenAI API |
| AutoCVE Security Audit Platform | 44 · Major gaps | Self-hosted serviceFree + model costs | ★ 1.2k | 1mo ago | Python | — |
| OrgKernel Agent Trust Layer | 66 · Some gaps | Library / SDKFree | ★ 2.7k | 3mo ago | Python | — |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence: API keys are stored via environment variables, no hardcoded credentials found; user must manually configure API keys, but no explicit user confirmation mechanism (e.g., second confirmation for dangerous operations) found; data flow is described in README architecture diagram, but detailed data flow and storage locations are not specified; sensitive data handling (e.g., Bilibili cookies) is mentioned but encryption measures are not detailed; dependency list includes known packages, but no vulnerability scan results provided; external effects include video download and upload to Bilibili (in development), but permission controls are not described; no rollback mechanism provided; source attribution: MIT license and copyright notice exist, but publisher is unverified. Deductions: lack of user confirmation, insufficient data flow transparency, missing sensitive data handling details, unverified dependency security, unclear permission controls for external effects, no rollback mechanism.
Evidence: README, SECURITY.md, CI config, and test files are largely consistent in feature descriptions, but some inconsistencies exist (e.g., version support in SECURITY.md vs README); dependency list is explicit, but no version pinning or integrity checks provided; error handling system has detailed tests, but no actual runtime verification. Deductions: dependency availability unverified, failure messages based only on code review, not actual execution.
Evidence: README clearly identifies target users (video creators, content teams) and use cases (YouTube/Bilibili video clipping); capability boundaries are described in feature list and in-development features; trigger precision: user triggers via UI, but trigger conditions are not detailed; environment fit: Docker and local deployment provided, supports macOS/Linux/Windows, but not all platforms are fully tested. Deductions: trigger precision description insufficient, environment fit not fully verified.
Evidence: README is well-structured with project structure, API docs, quick start; install notes are detailed including Docker and local deployment; naming stability: API endpoint naming is consistent, but version numbers are not explicit; examples and FAQ are rich; known limitations: in-development features are marked, but not all limitations are listed; license: MIT license file present; versioning changelog: no CHANGELOG provided; maintenance responsibility: SECURITY.md provides security contact email, but publisher is unverified. Deductions: naming stability not fully explicit, versioning changelog missing, maintenance responsibility not fully clear.
Evidence: Output usability: generated clips and collections are downloadable, but no actual output examples provided; marginal value: AI auto-clipping and smart collections are differentiating features, but no comparison with other tools; cost-benefit: requires API keys and compute resources, but no performance benchmarks provided. Deductions: output examples missing, cost-benefit not quantified.
Evidence: Feature claims in README have some code support (e.g., pipeline steps), but implementation details are not provided; cross-source verification: relies solely on single repository, no external validation; fact-inference separation: in-development features are clearly marked, but some claims (e.g., AI analysis capability) lack evidence. Deductions: claim traceability insufficient, cross-source corroboration missing, fact-inference separation not fully adequate.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Publisher identity is unverified; treat security claims with caution.
- Dependencies are not version-pinned, posing supply chain risks.
- Sensitive data handling (e.g., Bilibili cookies) lacks encryption details.
- No rollback mechanism; mistakes may be irreversible.
- AI features rely on external APIs; consider data privacy and costs.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
v1.5.6 AutoClip Desktop v1.5.6
Release notes excerpt
AutoClip Desktop v1.5.6 - 默认自动预检失败时改走字幕路线继续出片，不再整单失败；只有明确要求视觉分析时才停住。 - 导入前检查分析模型、转写、视觉和 FFmpeg。缺什么直接给出连接、安装或打开设置的入口；没有字幕且转写未就绪时不能开始。 - 自动出片失败事件补齐失败阶段、HTTP 状态和字幕/视觉路线，不再把所有失败都记成笼统的模型错误。 - 安装冒烟改为走 Studio 一键出片，并用 ffprobe 检查成片。 - 桌面版不再启动 Celery。项目列表以 studio.json 的出片终态为准，避免首页一直停在「制作中」。 - 删除已不再使用的旧导入页和未引用的 Celery 试验文件。 - 失败记录保留模型服务返回的 HTTP 状态；健康检查接口报告实际版本号；视频网站拒绝下载（HTTP 403/412/429）时显示中文原因和换网络或使用浏览器登录信息重试的建议，不再显示下载工具的英文原文。（#297） - 转写模型改为后台下载并显示真实状态，不再提前提示“下载完成”；导入前检查所选模型是否已在本地，缺少时可直接下载。模型下载默认先连官方源，失败后改用镜像。两个成片草稿不再选中同一段内容；成片渲染完成前首页不再显示“已完成”。（#298） …View on GitHub ↗ -
v1.5.5 AutoClip Desktop v1.5.5
Release notes excerpt
AutoClip Desktop v1.5.5 - 包含未公开发布的 1.5.4 全部修复（1.5.4 标签的发布构建被验收门禁误拦，未生成安装包），并修复发布构建读取验收标签的问题。 - 云端转写失败时保留云端原因和配置提示，不再误报本地 Whisper 未安装；保留原有请求上限。 - 成片保存后继续同步本次封面收尾，修复最后一张卡片封面生成后未自动显示的问题；可选封面失败或服务重启仍保留已生成视频。 - 发布简介中的 CEO、CTO、CFO、COO 或“公司负责人”等任职说法缺少原字幕依据时，保留标题与话题并省略简介，避免补入没有依据的身份或曾任经历；不增加模型请求。 - 静音切点检测兼容 Windows 本地编码和缺失的 FFmpeg 错误输出；可选停顿检测失败时仍可按字幕完成选段。 - 修复制作偏好和调整方案弹窗中下拉菜单被遮挡、无法用鼠标选择的问题。 - 修复隐私设置读取较慢时覆盖刚修改的崩溃报告开关，以及离开设置页后旧请求覆盖新选择的问题。 - 将访谈式成片的重点词标签和人物名牌放到字幕区域下方，避免与生成字幕或原视频字幕重叠。 - 模型目录的本地缓存无法写入时，继续返回已取得的账号模型列表；避免可选缓存故障阻断设置页刷新。 …View on GitHub ↗ -
v1.5.0 AutoClip Desktop v1.5.0
FAQ
Does AutoClip require a paid API key?
API_DASHSCOPE_API_KEY in the .env file. Without it, video downloading and basic processing still work, but AI highlight extraction will not function.What should I do if YouTube download fails?
pip install --upgrade yt-dlp. If the problem persists, try using browser cookies when creating the project.Does Bilibili download require login?
Can I deploy this system in production?
Can AutoClip work with other AI models?
API_MODEL_NAME). However, the project structure includes an LLM manager (llm_manager.py), so it might be extensible, but documentation does not provide guidance for adding other models.