Project · 2025

ai-audio-digest — Hebrew Audio & Video Transcription

An AI-powered audio and video transcription service specializing in Hebrew. Accepts files or URLs, transcribes using Whisper, and runs NLP for summarization and full-text search. Built with Lovable + GitHub.

Year 2025
Code →
Sound wave visualization with Hebrew transcription output

Tech

TypeScript Vite React shadcn-ui Whisper Lovable

The need grew out of personal frustration: I wanted to transcribe Hebrew videos and podcasts without paying for ChatGPT Premium every month or routing files through Otter.ai, which doesn’t handle Hebrew particularly well. ai-audio-digest accepts a file or URL, transcribes it, and returns clean text that you can search, summarize, or copy.

The project was built using the hybrid workflow I’ve adopted across all my recent personal projects: I started in Lovable to rapidly design the UI with shadcn-ui, then moved to a local IDE to write the serious logic — API calls, state management, and Whisper integration. Changes made in Lovable flow directly to GitHub and vice versa, giving me the freedom to move fluidly between both worlds.

The most interesting technical aspect was, in fact, the most minimal one — Whisper handles the transcription, so the real work was building a UI that doesn’t fall apart when dealing with a long file containing hundreds of lines of transcript output.

Outcomes

  • High-quality Hebrew transcription — including numbers, proper nouns, and punctuation
  • Clean interface built entirely with shadcn-ui and lucide-react icons
  • Hybrid workflow — Lovable for rapid UI design, GitHub for serious development work