This is a submission for the Cloudflare AI Challenge.

What I Built

AI Meeting Summarization models are advanced tools that can automatically process and evaluate large quantities of transcribed audio meeting data, providing useful insights.

Here is a brief overview of this model:

Transcription: Speech-to-text or automatic transcription models transform audio meeting data into a text format that is easier to handle.
Summary: AI-driven summarization models enhance the process even more. They condense extensive discussions into succinct synopses, encompassing essential elements, resolutions, and results. Envision a virtual assistant who attentively captures and comprehends the entirety of your discussion, subsequently delivering a concise and condensed overview, sparing you the arduous task of sorting through extensive hours of meeting recordings.
Precision and Insight: These models are supported by cutting-edge cloudfare AI research. They can recognize pivotal moments, emphasize actionable tasks, and discern between discussions centered on addressing issues and discussions centered on achieving desired results.
Application: The teams can utilize these meeting summaries to help them in the development process. AI summary simplifies the process by recognizing trends, tracking project progress, and assuring stakeholder agreement.
To put it simply (pun intended), AI Meeting Summarization models enable developers and product teams to extract significant insights from audio content, thereby enhancing the efficiency of meetings and improving the quality of decision-making. Next time you find yourself overwhelmed with meeting notes, contemplate delegating the task to artificial intelligence! 🎙️🤖

Demo

URL Demo link

My Code

Journey

This is the first time I use Cloudflare (after my friends always talk about it). I browse through Worker Documentation to grasp understanding Worker. I browse through models and I decide to make API of my idea using Workers. After testing it using Postman, I feel amazed (wow this is cool). But, I see other submission make Front-end, then I decide to learn about Pages. Since, I am not very good at making Front-end from scratch, I use Vanilla HTML to make form to upload file and show the Worker AI result. But, when I connect between Pages and Worker, I constantly get CORS. I was frustrated because I consistently forgot how to fix this CORS thing. But, I finally managed to solve it (I'm proud of myself lol). Then I decide to add Questions generation, I tried playgrounds to test different Text Generation, and I learn new thing about "Role" which can alter the output template, cool stuff.

Then, I realize there is Pages Function, I hope to explore it for my next Cloudflare project. I also learn how to process .wav file for Whisper model to use (thanks to Cloudflare tutorial).

Multiple Models and/or Triple Task Types
The model I use:

• CF Whisper Model (audio-to-text)
• CF Bart Large CNN (summarization)
• CF mistral-7b (text generation)