zhouxiaoka/autoclip

What is AutoClip?

AutoClip is a desktop‑first application (with CLI, Docker and MCP ‑ a local micro‑service) that turns a long‑form video (a talk, interview, podcast, etc.) into short, platform‑ready clips. You paste a video URL, pick a target platform (Douyin, TikTok, YouTube Shorts, B‑站, etc.), and AutoClip automatically:

1. Transcribes the audio (using Whisper, SenseVoice, or any cloud‑based speech‑to‑text service you configure).
2. Analyzes the transcript with a large language model to pick the most engaging moments, generate titles, descriptions, hashtags and a short caption.
3. Cuts the video, adds subtitles, chooses a visual style (portrait interview or full‑screen “podcast” layout), and optionally creates an AI‑generated cover image.
4. Packages everything into a ZIP that contains the short clip, a thumbnail, and the ready‑to‑post copy.

All heavy lifting (cutting, rendering) runs on your own computer; only the optional AI services (LLM, speech‑to‑text, image generation) call external APIs, and you pay for those calls yourself.

---

Who might use it?

• Content creators who have long interview or lecture videos and need dozens of 15‑60 s clips for TikTok, Instagram Reels, YouTube Shorts, etc.
• Social‑media managers looking to automate the repetitive work of finding highlights, adding subtitles and writing platform‑specific copy.
• Developers / hobbyists who want a self‑hosted, scriptable pipeline (Docker, CLI, or MCP) that can be integrated into larger automation workflows or AI agents.

---

Key features (as described in the README)

Feature	What it does
One‑click clipping	Paste a video link, select a platform, and AutoClip produces up to 10 ready‑to‑post short clips automatically.
Platform‑aware formatting	Chooses portrait vs. landscape, subtitle language, and caption style based on the target platform (e.g., interview style for Douyin, full‑screen podcast style for TikTok).
Local rendering	Video cutting and encoding happen on your machine via FFmpeg – no cloud rendering costs.
Flexible transcription	Use existing subtitles, local Whisper/SenseVoice, or any cloud transcription service you configure.
LLM‑driven highlight selection & copy	Connect to OpenAI‑compatible APIs, Ollama, LM Studio, etc., to let a language model pick highlights and write titles, descriptions, hashtags.
AI‑generated cover images (optional)	Can call image‑generation models (e.g., Gemini, GPT‑Image) for thumbnails, but this is off by default.
Multiple deployment modes	• Desktop app (Tauri + React) for macOS Apple Silicon and Windows x64. • Docker compose for self‑hosted web service. • Pure CLI / MCP for batch jobs or integration with agents.
Cost transparency	The README shows real‑world cost examples (e.g., ¥0.09 for a 7.5 min output using qwen‑plus).
Privacy‑first	Only transcript and selected frames are sent to the cloud when you use remote models; everything else stays locally.
Open source & MIT licensed	Free to use, modify, and redistribute.

---

How to get started (quick summary)

1. Download the latest release for your OS (macOS .dmg or Windows .exe) – the installer bundles Python and FFmpeg.
2. Configure a model in the Settings panel – either an OpenAI‑compatible endpoint (provide API key) or a local Ollama/LM Studio server.
3. Paste a video URL, choose the platform and portrait style, and click Generate.
4. Review the automatically created clips, edit if needed, then download the ZIP or post directly from the app.

For headless use, install the CLI package (autoclip-mcp wheel) or spin up the Docker stack and call the REST API documented at http://localhost:8000/docs.

---

Why it matters

Creating short‑form video is a major bottleneck for creators; manually finding highlights, adding subtitles and writing platform‑specific copy can take hours per hour of source material. AutoClip bundles several AI capabilities (speech‑to‑text, LLM summarisation, optional image generation) into a single, locally‑run workflow, dramatically reducing that time and cost while keeping the creator’s data under their control.

---

Where to learn more

• Official site & demo videos – https://zhouxiaoka.github.io/autoclip_intro/
• Installation guide – docs/USER_INSTALLATION_GUIDE.md
• Model configuration – docs/MULTI_LLM_PROVIDER_GUIDE.md
• CLI / MCP reference – docs/CLI_AND_MCP.md
• Research note on what drives view counts – docs/research/2026-10-what-drives-views/article-zh.md
• Community – GitHub Discussions, Issues, and the Star History badge.

---

TL;DR

AutoClip is an open‑source, locally‑run tool that uses AI (speech‑to‑text + LLM) to automatically cut long videos into short, platform‑specific clips with subtitles, titles and thumbnails, saving creators hours of manual editing while keeping data private.