LLM – Automatic creation of minutes from voice data (Whisper, ChatGPT)
This time, I will talk about tools that use OpenAI’s Whisper and ChatGPT, which are very popular among LLMs (Large Language Models). This is a tool that automatically creates minutes from audio data such as MP3 and WAV, and I would like to briefly introduce the Beta version.
I am using this tool myself, and I can now reduce the time taken to create meeting minutes from over an hour to about 5-10 minutes.
1. Background
Creating meeting minutes can be a time-consuming and tedious task. Sometimes I wonder what it would be like to have a note taker in the meeting. There are probably more things to really focus on. . With this motivation in mind, I was thinking every day if I could use the audio data from the meeting (which needs to be recorded with an iPhone, etc.) to transcribe it.
However, with the Beta release of Whisper from OpenAI, I created a tool. Also, when I immediately tried using it in my work, I was able to reduce the minutes, which took more than an hour, to about 5 to 10 minutes, including rework, and it was very convenient, so I decided to publish it.
2. Minutes automatic creation tool
Although still in Beta, I have created a prototype of a meeting minutes maker using Whisper and ChatGPT from OpenAI, a company that has received a lot of attention in LLM.
This demo uses the Private mode of the Huggingface environment. However, when actually using it, it is possible to meet security requirements by using Cloud such as AWS from the viewpoint of security.
1. Login
2. Main page
a. Minutes creation image (Japanese)
- Drag the mp3 file to the top left of the screen
- Select “jp” in Language
- Press “Send” button
- On the right side of the screen, the sentences transcribed from the voice are displayed.
- Minutes will be created at the bottom right of the screen. (The sample is created in Markdown, and the items are also samples.)
b. Minutes creation image (English)
Let’s create English meeting minutes using a Japanese mp3 file.
- Drag the mp3 file to the top left of the screen
- Select “en” in Language
- Press “Send” button
- On the right side of the screen, the sentences transcribed from the voice are displayed.
- At the bottom right of the screen, the minutes will be created in English. (The sample is created in Markdown, and the items are also samples.)
3. Inquiry method
If you would like to actually try it, I will share the sample page.
Please contact us after filling out the necessary information from the inquiry form below.