I've been building Harku, a transcription tool for creators, students, and researchers. One recurring problem was obvious: people often need the text from a public YouTube video, but don't want to upload a file, create an account, or learn a full editor.
So I built a focused YouTube-to-text flow:
- paste a public YouTube URL or video ID
- receive a timestamped transcript
- export TXT, SRT, or VTT
- use 60+ languages without signing up
The difficult part was not the transcription model. It was reducing the first-run workflow while still handling Shorts links, youtu.be links, restricted videos, and clear fallbacks.
The tool is live here: https://harku.io/tools/youtube-to-text
I'm looking for feedback on transcript readability, export formats, and the languages people actually need.