Lecture-STT
Lecture-STT is a full-stack academic productivity platform that lets students and researchers transcribe lectures, extract text from PDF slides and images, and ask questions about the content through conversational AI — all in one unified tool.
The Problem
Students attending lectures, watching recordings, or working through dense PDF slide decks have no efficient way to search, review, or query that content. Manual note-taking is slow and incomplete; existing transcription tools don't integrate with document extraction or AI Q&A.
The Solution
A three-modality ingestion platform: audio files go through Whisper ASR for high-accuracy transcription; PDF slides are parsed with pdfjs-dist with Tesseract.js OCR fallback for scanned documents; images are processed directly through Tesseract.js. All extracted content feeds a Groq-powered (LLaMA 3.3 70B) conversational AI that can answer questions, summarize, or generate study guides from the material.
Key Features
Challenges & Learnings
Complex PDF layouts (multi-column, overlapping text boxes, embedded tables) defeated standard pdfjs-dist extraction. A heuristic layout analysis layer was added to detect column boundaries and reorder text blocks before stitching. Groq's API rate limits required request queuing for long transcripts submitted as context.
Business Impact
Reduces time spent manually reviewing lecture materials by 60–80% for students. Particularly valuable for accessibility (students with hearing impairments can get accurate transcripts) and for second-language learners who benefit from the searchable text format.
Future Direction
YouTube video URL ingestion, real-time live lecture transcription via WebSocket streaming, and spaced-repetition flashcard generation from transcripts.
Tech Stack
Quick Links
Want to see more of my work?
Back to Projects