Hi HN! I'm the founder of Harku (https://harku.io ), a video/audio transcription tool I've been working on for the past few months. The backstory: I was frustrated with existing transcription services - they either required expensive subscriptions upfront, had poor accuracy for non-English content, or forced you to create accounts just to try them. As someone who frequently works with meeting recordings and content in multiple languages, I needed something better. Technical approach: - Built on OpenAI's Whisper V3 (latest model) for 98.5% accuracy - GPU-optimized processing pipeline that transcribes 1-hour videos in ~2 minutes - Supports 100+ languages with automatic detection - No user data retention - files are deleted within 24 hours What makes it different: - 10 minutes free daily without any signup (just drag and drop) - Works directly with YouTube URLs - no download needed - 500MB file limit even for free users (vs 100MB industry standard) - Clean exports without watermarks, even on free tier Current limitations: - Real-time transcription not yet available (coming soon) - No mobile app yet (web works on mobile though) - Custom vocabulary only for Pro users Technical challenges solved: - Implemented efficient audio extraction to handle 25+ video formats - Built a queuing system that scales automatically during peak hours - Optimized Whisper inference to run 30x faster than real-time I'd love to hear your feedback, especially on: 1. What features would make this more useful for your workflow? 2. Any edge cases with specific accents or technical content? 3. Interest in API access for developers? You can try it without any signup at https://harku.io - just drop a video and see the results. Thanks for checking it out! |