Case Study

Clipper

Self-hosted

AI-powered tool that automatically extracts viral moments from YouTube videos and generates TikTok-ready clips with burned-in Indonesian subtitles — speaker detection, auto-crop, full HD.

Open Source Free

Overview

Short-form content is winning, but cutting clips from long videos is tedious. Clipper automates the full pipeline — from a YouTube URL to a TikTok-ready vertical clip with burned-in Indonesian subtitles.

The Problem

Content creators spend hours scrubbing through videos to find the right moments, then more time cropping, captioning, and exporting. For Indonesian-language content, subtitle tooling is even more limited.

What Was Built

A full-stack tool (FastAPI + SvelteKit) with a six-step pipeline:

  • Download — fetch video at 1080p via pytubefix
  • Transcribe — extract audio and run speech-to-text via Whisper
  • Analyze — Gemini scores segments for virality (0–100) and picks the best moments
  • Clip — FFmpeg cuts and burns in subtitles
  • Format — output in 9:16 (TikTok) or 16:9 (YouTube), with MediaPipe-powered speaker tracking and smooth auto-crop
  • Cleanup — source files auto-delete after export

Technical Notes

Supports up to 10 clips per run, configurable Whisper model size (tiny → large), and full HD output up to 1080p.


View Project All projects →