Spaces:
Sleeping
Sleeping
| title: Lecture Processor | |
| emoji: "\U0001F393" | |
| colorFrom: blue | |
| colorTo: indigo | |
| sdk: gradio | |
| sdk_version: 5.15.0 | |
| app_file: app.py | |
| pinned: false | |
| license: mit | |
| hardware: zero-a10g | |
| # Lecture Processor | |
| Transcribe, summarize, and generate quizzes from lecture recordings using **WhisperX** and a fine-tuned **Gemma 3 4B** model. | |
| ## How It Works | |
| 1. Paste a YouTube lecture URL | |
| 2. The pipeline automatically: | |
| - **Transcribes** speech using WhisperX | |
| - **Summarizes** the lecture with structured sections (Summary, Key Points, Action Points) | |
| - **Generates quiz questions** (5 MCQ + 3 short answer) | |
| ## Tech Stack | |
| - **WhisperX** - Speech-to-text transcription | |
| - **Gemma 3 4B Instruct** - Fine-tuned with QLoRA for lecture summarization and quiz generation | |
| - **LoRA Adapter** - [noufwithy/gemma-lecture-adapter](https://huggingface.co/noufwithy/gemma-lecture-adapter) | |
| - **Gradio** - Web interface with ZeroGPU support | |