wtf is auto-clip?

This project is a local Python tool that automates turning a long YouTube video into a set of short clips formatted for platforms like TikTok, Instagram Reels, and YouTube Shorts. Everything runs on your own machine: the download, transcription, clip selection, cutting, subtitle adding, and export all happen locally without sending video content to an external server. The workflow starts by downloading the YouTube video using a tool called yt-dlp, then extracting its audio and transcribing it with OpenAI Whisper running locally. Whisper produces a text transcript and a subtitle file. An AI analyzer then reads the transcript and selects ten to twenty interesting moments to clip. By default the project tries AI providers in order, including Gemini, GPT, Claude, DeepSeek, Qwen, and others, falling back to a local rule-based analyzer if none of the API providers are available or configured. Each selected clip is cut to a vertical 9:16 aspect ratio at 1080 by 1920 resolution, with large white subtitles and a black outline placed at the center-bottom of the frame. The tool also applies a light automatic zoom effect, adds a short intro text from the clip's opening line, and optionally runs an automatic silence-removal pass using a separate tool called Auto-Editor. Finished clips land in an output/clips/ folder alongside a JSON metadata file for each clip. Running the tool on Windows requires Python 3.10 or 3.11, FFmpeg, and the packages listed in requirements.txt. Whisper model size is configurable: smaller models run faster on slower machines (tiny, base, small, medium, large in order from lightest to heaviest). API keys are optional since the local fallback analyzer can produce a clip plan without any external service. The README is written in Indonesian.