MeetScribe

GIF
MeetScribe Demo

Inspiration

The inspiration behind MeetScribe came from the need for an AI-powered assistant that can quickly inform teams about what was discussed, the actions to be taken, and the overall sentiment of the conversation.

What it does

MeetScribe listens to your meeting recordings and provides all the essential details and takeaways you need—summaries, action items, and sentiment analysis, so you never miss out anything.

How we built it

We used pre-trained models from OpenAI and Hugging Face for transcription, summarization, and sentiment analysis.
For transcribing the audio we used OpenAI’s Whisper model.
For identifying next steps and action items, we leveraged the spaCy NLP library.
The front end was built using Streamlit, which allowed for seamless integration with our backend code running on Google Collab.

Challenges we ran into

One of the main challenges was integrating our backend built in flask with the initial front end, which was built using HTML, CSS, and JavaScript.
We were having the trouble running the application in GPU. We would have go for FastAPI but it would have also taken more time than we had.
A mentor suggested using Streamlit, which proved to be a game-changer due to its compatibility with Colab and ease of use. Switching to Streamlit at the last minute required us to integrate our initial UI design into the backend python script in streamlit for all of our deliverables.

Accomplishments that we're proud of

We're proud of how quickly we adapted to the unexpected switch in our tech stack. Despite the last-minute change, we were able to deliver a fully functional product, and that adaptability and teamwork made all the difference.

What we learned

We learned how to work with Hugging Face Transformers, OpenAI's Whisper for transcription, front end development with Streamlit, and backend–frontend integration.
Additionally, beyond the technical skills, we learned that all it takes is one focused weekend to build a working prototype of a cool idea.

What's next for MeetScribe

We would like to host MeetScribe on a cloud platform and build a browser extension that can listen to meetings live and provide real-time insights and takeaways along with complete multilingual support.

GitHub Link: https://github.com/sudswain-hue/AI-Powered-Meeting-Companion/tree/main