Transcript and Summarize:

Recording long conversation, trancipting, then summarize, has became much simpler and affordable with new LLMs.

This is a small project to utilize OpenAI and Speech recognizer (supports various models) APIs.

Using the user primary microphone and a streamlit web UI, the user could choose one of three languages [English, German, Arabic], total duration of the recording time in minutes, and how many seconds to save as a file as the time goes.

Steps to Run

1. Clone Repository

git clone https://github.com/ahmadhatahet/transcript_and_summarize.git
cd transcript_and_summarize

2. Create virtual environment

python -m venv transcript_and_summarize

# windows
source transcript_and_summarize\Scripts\activate

# linux
# source transcript_and_summarize/bin/activate

3. Install packages

pip install -r requirements.txt

4. Start Streamlit app

streamlit run app.py

Todos:

1. ✔ Multi language support (EN, DE, AR).
2. ✔ Add options to the UI for faster modifications.

Notes:

Arabic language is one of the Semitic language , thus, it is not well recognized by the models used yet. Transcripting works fairly well, but summarizing is less than average. Maybe a better prompt could help increase the quality of the summarized text.

Screenshot: