LectureMind: Voice AI Lecture Synthesizer

In higher education, university students attend 60-to-90 minute lectures filled with dense academic terminology and complex formulas. Traditional note-taking forces an overwhelming cognitive load: students either focus on writing verbatim notes or listening to comprehend concepts—rarely both. Existing transcription tools fail because they hallucinate technical vocabulary, produce unformatted walls of text, and drop 70–80% of spoken technical content. LectureMind (Apna Lecture, Apni Zubaan) is an autonomous, voice-first academic knowledge synthesis studio engineered for university scholars (18+). It captures live or recorded lecture speech, transcribes it with sub-second latency via AssemblyAI's Universal-2 STT engine, and transforms 100% of the speech into structured revision assets: 1. Interactive 4-Tier Concept Mind Maps: A visual knowledge graph mapping Root -> Thematic Pillars -> Sub-concepts -> Mechanism & Evidence Leaves, along with golden Domain Expansion nodes and 1-tap JPG export. 2. Structured Academic Revision Notes: Dense Markdown summaries containing comprehensive pedagogical explanations, core technical definitions, and exam takeaways. 3. Psychometric Socratic Quizzes: Formative assessments grounded strictly in Bloom's Taxonomy with instant explanatory feedback for active recall. 4. Bilingual Voice Intelligence: Seamless native code-switching support between English and Urdu. Architecturally, Lecture Mind leverages Assembly AI streaming STT paired with dual-LLM orchestration (Groq LLaMA 3.3 70B reasoning and an intelligent on-device synthesizer for zero downtime). Built with Flutter Web, it delivers a Dark Navy Neumorphic UI and is deployed live on GitHub Pages.