ReCapAI

• 

• 

  recap.docx

• 

  logo

• 

  prints to terminal too!

• 

Inspiration

An audio recapper/analyzer tutorial is provided on the openai website. The key difference for this project is that gradio is being used as an interface so that anyone in the public can access this via web link

What it does

Recaps into 1 paragraph, analyzes tone, and extracts key points & action items of transcribed audio file so that users can quickly understand the main points and focus that was discussed in an audio file without having to listen to it

How we built it

This project would not be successful without the help of packages such as openai, whispers, and gradio.

Challenges we ran into

• interacting with google colab
• whispers can't transcribe audio files over 25MB, thus, needed to figure out how to chunk audio files into smaller parts then concatenate them back together for OpenAI chat completion
• there were dependencies that needed to be upgraded before pip install gradio & import gradio could be successful
• Difficult to get gradio to display the recap.docx file that was generated

Accomplishments that we're proud of

• On the backend, chatgpt successfully generates a coherent, concise, and non-hallucinating recap of the audio
• On the backend, chatgpt's completion was successfully saved into a Word document

What we learned

• chatCompletions and how chatGPT API works
• quickly create a professional UI experience with gradio

What's next for ReCapAI

• Improve the UI using gradio so that users can upload their own audio file download the recap.docx file that is generated
• Allow for users to upload bigger files

Built With

• assemblyai
• gradio
• openai
• python