muapi-ai-clipping

muapi-ai-clipping turns a long video into a set of short, vertical, viral-ready clips through a single managed API call. It solves the heavy local-tooling problem of short-form clipping: instead of running Whisper for transcription, an LLM for highlight ranking, ffmpeg for cropping, and a GPU for all of it, the skill wraps muapi.ai's /ai-clipping endpoint, which performs the entire pipeline server-side. Each returned clip carries a viral score (0-100), an opening hook line, a one-sentence reason it works, and a hosted mp4 URL.

The skill is a thin command-line wrapper. An agent execution protocol collects inputs (a hosted mp4 URL, local file, or YouTube URL, plus optional number of clips, aspect ratio, and a coords-only mode), verifies that the muapi CLI is installed and authenticated and that an API key is available, then runs a bundled bash script. The script resolves the video to a hosted URL (uploading local files through the CLI when needed), calls the clipping endpoint, polls until the job completes or returns a request_id under async mode, and prints a ranked summary. Server-side the endpoint transcribes, classifies content type, ranks highlights through a virality framework (hooks, emotional peaks, opinion bombs, revelations, conflict, quotable lines, story peaks, practical value), dedupes overlapping candidates, selects the top N, and face-tracks an auto-crop to 9:16, 1:1, or 4:5. The documentation includes invocation patterns, an aspect-ratio picker, an output schema, common mistakes, and failure modes.

The target users are content creators, podcasters, and social teams repurposing podcasts, interviews, lectures, vlogs, or streams into TikTok, Reels, and Shorts without a local media pipeline. Note that using it uploads video content and requires a muapi API key; all heavy processing and hosting happen on the third-party service by design.