pavithramurugan824-oss/lecturesync-ai

LectureSync is a documentation-first repository: it defines the product, pipeline, and system design for an AI workflow that turns lecture videos into study notes and quizzes. It is not presenting an implemented application here; the architecture is described across the Markdown docs.

The core system boundary is simple: a user provides either an uploaded MP4 lecture or a YouTube lecture link, and the system returns two artifacts — a summarized PDF of study notes and a generated 10-question quiz.

The main pipeline is linear and content-processing oriented. First comes ingestion of the lecture source, including handling two input modes: direct video upload and remote video retrieval from YouTube. That creates the raw media asset that feeds the rest of the system.

Next is audio extraction and transcription. The lecture audio is converted into text using speech-to-text processing, which becomes the canonical intermediate representation for all downstream AI steps.

After transcription, the AI pipeline performs concept extraction. This stage identifies key topics, learning objectives, and salient points from the transcript, separating signal from filler so later outputs stay focused.

The study-note generator is one output branch from the extracted content. It composes a structured summary suitable for revision, then exports it as a PDF. In architecture terms, this is a document-generation boundary fed by the transcript/concept layer.

The quiz generator is the second output branch. It uses the lecture content to produce an assessment aligned to the material, specifically a 10-question quiz. This is a distinct AI subdomain from summarization because it must preserve coverage, difficulty balance, and content specificity.

The repository’s docs suggest a modular pipeline rather than a monolith: ingestion, transcription, concept extraction, note generation, quiz generation, and export are separate concerns. That separation is important for the architecture diagram because each stage has its own data contract and failure mode.

The docs/ai-pipeline.md, docs/quiz-generation.md, and docs/system-architecture.md files likely describe the processing flow and component boundaries, while docs/demo/demo-flow.md captures the end-to-end user journey. docs/problem-statement.md frames the use case, and tech-stack.md captures the intended implementation stack.

From a systems perspective, the most important data flow is: video source → audio/transcript → concepts → two parallel outputs. The transcript is the central shared artifact, and the note/quiz generators are downstream consumers of that shared semantic layer.

For an architecture diagram, the clearest boundaries are:

• User input layer

• Media acquisition layer

• Speech-to-text layer

• AI analysis layer

• Study-note generation/export layer

• Quiz generation layer

Overall, LectureSync is a single-purpose AI content pipeline optimized for educational content transformation, with a strong emphasis on automated post-processing of lecture recordings into reusable learning assets.