Show HN: Transcriptum – fast video transcription with speaker labels and summary

Transcriptum is an audio and video transcription service powered by WhisperX: speaker diarization, word-level timestamps, 50+ languages, and optional AI analysis (summaries, Q&A, topics, sentiment, action items, fact-checking) via multiple LLM providers (OpenAI, Gemini, DeepSeek).

You can upload files or paste a YouTube URL; we download and transcribe. Export as TXT, SRT, VTT, DOCX. Built with NestJS, Next.js, Prisma/PostgreSQL; subscriptions are handled with Polar.

I built this to have accurate, private transcription without locking into a single vendor, and to combine transcription with analysis in one place. Would love feedback from anyone who transcribes meetings, podcasts, or long-form content.