MeetMate: AI-powered virtual meeting assistant

MeetMate is an AI assistant that provides transcripts and summaries for online meetings. Upload your meeting recording for speaker-wise notes, minutes, and action items in just one click!

Comment

Inspiration

Our main inspiration for this project is the time-consuming task of creating Meeting Minutes by hand after every virtual meeting. As college students who are part of several student chapters, we felt that there is a serious need to automate this tedious yet crucial job, and we realised the potential AI can bring into this field. Thus MeetMate was born, the perfect AI companion for all your meeting summarisation tasks.

What it does

MeetMate is an AI-powered virtual meeting assistant that lets users quickly generate detailed summaries and documents about their meetings. Just upload the meeting recording, and along with the meeting Minutes, MeetMate also provides a full audio transcript and automatically schedules follow-up meetings in the user's Google Calendar.

How we built it

We built MeetMate in Python and utilised Microsoft's Azure AI Foundry, Azure Speech Services, and their supporting SDK packages. We built the audio transcription and diarization with Azure AI Speech's Fast Transcription REST API and used OpenAI's GPT-4o for Minutes generation and document creation. MeetMate also uses the Google Cloud Platform (GCP) to access the Google Calendar API for automated event scheduling. All this sweet functionality is under the hood of a slick webpage UI built using HTML5, CSS and Flask. During the entire dev cycle, we used GitHub Copilot as our AI pair programmer to speed up the coding phase, generate code snippets, and fix bugs.

Challenges we ran into

We ran into several challenges while configuring the AI Foundry project with the correct endpoints and API keys. Luckily with excellent documentation, we were able to clear them all out. Also, we faced issues with using Azure OpenAI for video transcription via Whisper and Minutes generation via GPT-4o. Since Azure OpenAI services were accessible only to Azure Enterprise accounts, which were out of our team's expense limits, we resorted to Azure Fast Transcription and external OpenAI API key methods.

Accomplishments that we're proud of

We're very proud of how seamlessly MeetMate works, working hand-in-hand to generate excellent results from audio recordings. We're especially happy with how quickly we were able to configure the Fast Transcription REST API, which we first tested with several parameters in the Azure AI Playgrounds. We're also proud of the follow-up scheduling automation we made using Google Calendar API. We're also quite happy with the MeetMate UI, offering a simple, accessible and clean look to it.

What we learned

We learned that AI features and cloud platforms like Azure AI Foundry, OpenAI and Google Cloud are extremely essential tools for developers like us to master. And we greatly extended our Python and web design proficiencies through this dev cycle. We also learned by experience how to efficiently and quickly get started with development by using GitHub Copilot. We're very happy to have explored and implemented the use of AI in online meeting transcriptions and have learned that this knowledge of generative AI can be fitted to solve all the world's problems one day.

What's next for MeetMate: AI-powered virtual meeting assistant

MeetMate has a list of features that are currently under development or in ideation:

  • MeetMate as a Chrome extension; for direct in-meeting summarisation and closed captions.
  • MeetMate as a Microsoft Teams feature; for easy availability of previous recordings' notes.
  • Support for audio translation to and from multiple languages like German, Spanish, Hindi, Tamil, etc.; for users in regions of high linguistic and cultural diversities.

Built With

Share this project:

Updates