TranscriptMax
Project info
- Second-accurate transcription from YouTube, Instagram & TikTok
- AI summary, key points and automatic chapter segmentation
- Speaker diarization: who says what, throughout the video
- Timestamp-anchored annotations + global search across all your transcripts
- Multi-format export (PDF, Markdown, SRT, Notion, Obsidian, Readwise) and 40+ languages
- FastAPI + PostgreSQL/Redis backend, Next.js 14 frontend, 4 dockerized services
About the project
TranscriptMax turns any YouTube, Instagram or TikTok video into a clean, second-accurate transcript. The idea: stop watching hours of video and instead read, annotate and archive it like real documents.
Every transcript comes with an AI-generated summary and key points, automatic chapter segmentation and speaker diarization (who says what). You highlight and annotate like in a book — each note stays anchored to its timestamp — search a word across all of your transcripts, and export to PDF, Markdown, SRT, Notion, Obsidian or Readwise. A transcript can also be shared read-only via a public link.
Under the hood the product runs on a FastAPI (Python) backend with PostgreSQL for data and Redis for caching, JWT authentication, and a Next.js 14 frontend (App Router, TypeScript, Tailwind). Everything is containerized: four services orchestrated with Docker Compose (frontend, backend, postgres, redis), plus a companion browser extension.
Gallery
