Hey HN,

  I built an open source tool to transcribe Instagram Reels using OpenAI's Whisper model.

  GitHub: https://github.com/stym06/reelstudio

  How it works:
  - Paste a Reel URL → Worker downloads via yt-dlp → Whisper transcribes → Get shareable transcript page

  Tech stack:
  - Next.js 14 frontend
  - Go worker (polls PostgreSQL for jobs)
  - faster-whisper for transcription
  - No login required

  The architecture is simple: the frontend creates a task in PostgreSQL, a Go worker polls for pending tasks, downloads the video, runs Whisper, and saves the transcript. The
  frontend polls until complete and redirects to the transcript page.

  I built this because I wanted to repurpose Reel content into blog posts and needed accurate transcripts with timestamps. Existing tools either required accounts, had watermarks, or
   were expensive.

  Self-hosting is straightforward - just need PostgreSQL, the Next.js server, and the Go worker with Python/Whisper installed.

  Would love feedback on the architecture or any feature suggestions. The worker is designed to be horizontally scalable if anyone wants to add Redis/queue support.