Technical highlights: • 99%+ accuracy using Whisper with optimized chunking for long audio • Speaker diarization without training data • AI chapter detection using topic boundary analysis • Exports to Markdown/TXT/JSON with structured metadata
Free tier: 5-minute preview Tech stack: Next.js, Whisper, R2 storage
Try it: https://harku.io/tools/podcast-transcription
Built this after spending 30-60 minutes manually creating Show Notes for each episode. Would love feedback from fellow podcasters and developers.