Noter 👩🏻‍🏫

Noter turns lecture audio into detailed summary sheets.

Features

• Transcribe lectures: Whether it's a live recording or prerecorded, Noter can transcribe it into a .txt file.
• Summarize lectures: Noter uses GPT-4o to create (.md) files with summaries, definitions, and step-by-step examples covered in a lecture. Summary sheets can be created from all modern video or audio formats or a transcript (.txt) file.
• Playback lectures: Relisten to lectures with the (.mp3) files automatically sorted by class code and date.

Quick Start

First setup repo:

MacOS users with M1 chip or newer follow here

Other operating systems are not currently supported!

Then use repo example:

Learn how to summarize from transcript here

Demo

Press on the thumbnail to watch!

Installation

macOS (M1 or later) 🍎

Prerequisites 📋

• Git
• Python 3.10 (miniconda recommended)
• ffmpeg

If you don't already have these installed here's how you can do so!

1. Install Homebrew

   /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

2. Install Git

   brew install git

3. Install miniconda (strongly recommended to avoid python version conflicts)

   Download miniconda

   mkdir -p ~/miniconda3
   curl https://repo.anaconda.com/miniconda/Miniconda3-latest-MacOSX-arm64.sh -o ~/miniconda3/miniconda.sh
   bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3
   rm -rf ~/miniconda3/miniconda.sh

   Initialize miniconda

   ~/miniconda3/bin/conda init zsh

4. Install ffmpeg

   brew install ffmpeg

Setting up this repo 😇

1. Clone the Repository:

   Clone in the directory you want this project in!

   git clone https://github.com/jadenScali/noter.git

   Move into this project

   cd noter

2. Create and activate a conda Environment (Optional) (Recommended):

   Python 3.10 is required for this project. You may install it manually but it may cause python versioning conflicts. We strongly recommend you use miniconda.

   conda create -n noter python=3.10
   conda activate noter

3. Create Whisper (speech to text) model locally:

   Install requirements and make model

   pip install -r requirements.txt
   git clone https://github.com/ggerganov/whisper.cpp.git
   cd whisper.cpp
   make medium.en

   Install coreML support for faster runtimes on M-series chips

   To ensure coremltools operates correctly, please confirm that Xcode is installed and execute xcode-select --install to install the command-line tools.

   ./models/generate-coreml-model.sh medium.en
   make clean
   WHISPER_COREML=1 make -j

   Cleanup folders

   cd ..
   mkdir -p whisper/models
   mv whisper.cpp/main whisper/
   mv whisper.cpp/models/coreml-encoder-medium.en.mlpackage whisper/models
   mv whisper.cpp/models/ggml-medium.en-encoder.mlmodelc whisper/models
   mv whisper.cpp/models/ggml-medium.en.bin whisper/models
   rm -rf whisper.cpp

   NOTE: The first time the speech-to-text model is ran it will be slow because it's optimising for your hardware

4. Setup AI features using OpenAI

   Create an openAI account and follow these instructions to generate an API key

   Create a .env file in the root of the project directory:

   touch .env

   Open the .env file in a text editor and add your OpenAI API key:

   OPENAI_API_KEY="your_openai_api_key_here"

Example Usage

1. Start program (make sure your conda environment is active if you used conda)

   python main.py

2. Follow the instructions and create your classes

3. Select 3 (summarise from transcript)

4. Use the samples/cs50_SQL_transcript.txt to summarise the sample transcript

5. Fill out the remaining fields

6. Note for Lecture # you can use decimal numbers especially if you want to create multiple summaries for the same transcript

7. After the summary sheet is created you can check it under the summaries folder at the path notes/COURSE_CODE/summaries

8. To properly view the .md file use a Markdown viewer in or out of your IDE. If you're using vsCode there are good extensions you should use to properly display the Latex math symbols