autoclip/README-EN.md

AutoClip

Open-source AI highlight clipping

Turn long videos into moments worth sharing.

Download desktop app · Quick start · Website · Documentation · Report an issue

简体中文 · English · 日本語 · 한국어 · Español · Português · Русский · Français

AutoClip uses AI to analyze video transcripts, find highlights, write titles, and create clips and collections. Built for interviews, podcasts, courses, and livestream recordings, it offers a desktop app, a Docker web interface, and CLI / MCP access.

See the interface

AI-generated clips	Studio preview and editing

_{Real v1.4.0 UI with subsequent Studio fixes: import videos, review actual generated clips, and edit in Studio. The UI is in Chinese; the sample transcript and generated titles remain in English.}

Screenshot version and sample source (Chinese)

Special thanks ❤️

Infistar.cc

Thank you to Infistar.cc for sponsoring AutoClip! Its multi-model API service can support long-video transcript analysis, highlight selection, and title generation. ⚙️ Compatible setup: Choose the OpenAI-compatible provider in AutoClip and enter the Base URL, your API key, and an available model. 🧩 Model choice: The partner offers Claude, GPT, Gemini, DeepSeek, and other model families. Choose models that support the compatible endpoint to compare transcript analysis and highlight selection. 🏷️ Pricing and services: According to the partner, selected models cost as little as 1% of official list prices, with RMB billing, invoices, and model authenticity verification. Check the platform for eligible models, current prices, and service terms. 🎁 AutoClip offer: New users can receive $5 in trial credit through our referral registration link, subject to the promotion terms. Setup guide

What you can do

Click any thumbnail to view the full image.

Import footage Use local videos, YouTube or Bilibili links, with optional SRT subtitles.	Find highlights Extract outlines, topic timelines, highlight scores, and clip titles from transcripts.
Create clips and collections Generate clips and suggested collections, then adjust their order manually.	Export for publishing Use presets for Douyin, Xiaohongshu, YouTube Shorts, and Bilibili, with burned-in subtitles and title cards.
Covers and publishing Since v1.3.2, generate covers and publish immediately or on a schedule. Connect overseas platforms through your Upload-Post account; configure Bilibili separately. No publishing account is connected in this demo. These images show the publishing entry and cover settings, not completed posts.	Publishing management Review publishing history and the calendar, manage pending posts, or simply download your clips.
Choose your models Qwen, OpenAI-compatible APIs, Gemini and other cloud services, or local models through Ollama / LM Studio.	Automate your workflow Orchestrate runs with the CLI or call the same processing pipeline from an MCP client. CLI / MCP has no GUI: this image captures a display page of actual command help output.
Multilingual interface Since v1.3.1, the app, website and README support Chinese, English, Japanese, Korean, Spanish, Portuguese, Russian and French. Choose a language in the header or follow your system. Your media and generated content keep their original language. English interface and language menu; media and generated content retain their original language.

Platforms, account requirements, and export details

After clips are ready, open Publish on a clip. Available in v1.3.2. Overseas publishing uses platforms connected on your own Upload-Post account: TikTok, Instagram, YouTube, Facebook, LinkedIn, X, Threads, Pinterest, Bluesky, Discord, Telegram, and Google Business, as available on that account. Bilibili is one account: paste a Cookie once in Settings.

It must include SESSDATA, bili_jct, and DedeUserID. Publish now or on a schedule. Title and description are optional and default to the clip title. Burned-in captions default on, and the ~4s title card defaults on.

Visibility defaults to private / self where the platform supports it. AutoClip promises that only for TikTok, YouTube, and Bilibili. You can download without publishing. The project page shows publish history and a calendar, and can cancel a schedule that has not gone out.

“Plan this week” fills Monday, Wednesday, and Friday at 09:00 for overseas platforms only, not Bilibili. Vertical accounts render 9:16 without a 60-second cut. Bilibili alone renders landscape. LinkedIn or X alone keeps the original frame.

Vertical and Bilibili in the same batch are rendered separately.

When publishing, a cover can be generated automatically so Bilibili does not reject an empty cover. Default cover and title-card details follow that release’s installer notes. Available in v1.3.2.

Import video → Subtitles / transcription → AI analysis and scoring → Clips and collections → Export

Quick start

Your workflow	Recommended option	Requirements
Edit on your computer	Desktop app	macOS Apple Silicon / Windows x64
Self-host / Linux	Docker	Docker + Compose v2
Batch processing / agents	CLI / MCP	Python 3.10+ (3.11 recommended) + FFmpeg

Desktop: your first clips

1. Install. Download the appropriate installer from Releases: .dmg for macOS Apple Silicon or -setup.exe for Windows 10 / 11 x64. Python and FFmpeg are bundled. Intel Mac / Linux users can use Docker or the CLI. Check the release for system requirements.
2. Configure a model. Select a provider in Settings, enter your API key and model, test the connection, and save. For local models, start Ollama or LM Studio first.
3. Import a video. Start with a 3–5 minute sample, optionally with SRT subtitles. Without subtitles, prepare the local Whisper components and speech model in Settings first.
4. Preview and export. Check clip boundaries, titles, and content, then select an export preset or connect an account to publish.

Full installation guide · Troubleshooting

Docker / Web

Requires Docker and Docker Compose v2. Run these commands from the repository root:

git clone https://github.com/zhouxiaoka/autoclip.git
cd autoclip

cp env.example .env

Before starting, edit .env: select LLM_PROVIDER and set the matching API key and model name. You can also configure the provider in Settings after startup.

mkdir -p data logs uploads
docker compose up -d --build

Open the web interface. API documentation is available once the backend starts. See the Docker guide for deployment details.

On Linux, if bind-mounted directories cause permission errors, fix ownership of the project data directories with this command, then start the services again:

docker compose run --rm --no-deps --user root --entrypoint sh autoclip -c 'chown -R autoclip:autoclip /app/data /app/logs /app/uploads'
docker compose up -d

CLI / MCP

Requires Python 3.10+ (3.11 recommended) and FFmpeg on PATH. The installation example uses a macOS / Linux shell; on Windows PowerShell, activate with venv\Scripts\Activate.ps1. Local CLI processing does not require Redis.

git clone https://github.com/zhouxiaoka/autoclip.git
cd autoclip
python3 -m venv venv
source venv/bin/activate
python -m pip install -r requirements.txt
python -m pip install -e .

Local model example: install and start Ollama, then download a model. Videos without subtitles require faster-whisper; the speech model is downloaded on first use. To supply existing subtitles, add --srt talk.srt.

ollama pull qwen2.5:7b
python -m pip install faster-whisper
autoclip doctor --provider ollama
autoclip run talk.mp4 --provider ollama --json

Replace PROJECT_ID with the project ID returned by processing to export for Shorts. Start the stdio MCP server with autoclip mcp:

autoclip export PROJECT_ID --preset shorts
autoclip mcp

In your MCP client, set command to the absolute path of autoclip in your virtual environment and args to ["mcp"]. See the CLI / MCP guide and Agent skill (both in Chinese).

Model configuration

Option	Configuration
Cloud models	Select a provider in Settings and enter your API key and model. OpenAI-compatible services also accept a custom Base URL.
Ollama	Default endpoint: http://localhost:11434/v1; default model: qwen2.5:7b. No API key required.
LM Studio	Load a model and start Local Server at http://localhost:1234/v1 by default. Select a model actually served by your instance.

Inside Docker, localhost refers to the container. To use a model on the host, configure an address reachable from the container; see the CLI / MCP guide. Video cutting runs locally; cloud model analysis sends transcript text to your selected provider. Video and model downloads still need internet access.

Infistar · Setup guide

Frequently asked questions

Is it free? Do I need an API key?

AutoClip itself stays free and open source under MIT. Cloud providers bill their own model usage and require your API key. Ollama / LM Studio presets need no cloud key, but require model files and suitable hardware. As of v1.3.2, overseas publishing needs your own Upload-Post account. Free and paid tiers, and daily caps for TikTok, YouTube, Instagram, and other platforms, follow Upload-Post’s own pages. They are not AutoClip promises.

Are my videos uploaded?

Editing stays on your device. Cloud model analysis sends transcript text to your selected provider. The finished clip leaves the machine only after you click Publish, and only to the platforms you connected. You can also download it without publishing. That Publish page is available in v1.3.2. Analytics and error reporting depend on your version and settings; see the privacy notes.

Can I use videos without subtitles?

Yes, after preparing local Whisper components and a speech model. You can also import existing SRT subtitles. Accurate subtitles can reduce transcription time and recognition errors.

Why were no clips generated?

Check the failed stage: empty subtitles, model connection failures, an overly high score threshold, or FFmpeg/disk problems. You can try reducing the threshold from 0.7 to 0.5, but this does not guarantee clips.

What videos work best? How long does it take?

Analysis primarily uses transcripts, making interviews, podcasts, lectures, and spoken commentary suitable. Purely visual action or music may work less well. Time depends on duration, hardware, models, and export settings; start with a 3–5 minute clip you provide.

See the first-clip guide for sample preparation and public video examples.

Full troubleshooting guide · Known issues

Documentation

Guide	Link
Getting started	Installation
Hosting and automation	Docker · CLI / MCP (Chinese) · Agent skill (Chinese)
Models and troubleshooting	Model configuration (Chinese) · Troubleshooting
Versions and privacy	Changelog · Privacy
Development and translation	Contributing (Chinese) · Translation maintenance (Chinese)
Partner setup	Infistar

The README is available in eight languages. Installation, Docker, and troubleshooting guides are available in English; other detailed references are mainly in Chinese.

Contribute and connect

Contributions, feedback, and translation improvements are welcome. For bug reports, include your OS, version, model, reproduction steps, and error logs with sensitive information removed.

Maintained by an individual in their spare time. Response times vary; live support and one-to-one deployment assistance are not provided. Please check the FAQ and known issues before contacting the maintainer.

Ideas, use cases, and model requests belong in GitHub Discussions. Reproducible bugs use the issue form. Board rules: community board (Chinese).

• Welcome and categories

• First-clip Q&A

• Ideas

• Email: christine_zhouye@163.com

Thanks to FastAPI, React, Tauri, FFmpeg, yt-dlp, Whisper, and all contributors. Licensed under the MIT License. If AutoClip helps you, consider giving the project a star.

Community recognition · Star History

These badges are provided by Trendshift. Click to view AutoClip’s recorded achievements. GitHub Trending and Trendshift are separate rankings; badges show recorded achievements, not a live position.