Automatic zoom reports

A tool for transcribing meeting text and summarizing it

This tool allows you to submit a recording of a meeting and receive a structured report on it.

A report contains:

• keywords,
• short summary of the meeting,
• structured summary by the topics,
• transcript, divided by speakers.

Here you can see an example of a report.

This example was created based on the video of the career panel of Data Fest Siberia 3.

How to use

1. Clone the repository, install requirements:

git clone https://github.com/Aillustrate/automatic_zoom_reports
bash install.sh

2. For the tool to work, you need to specify two API tokens in the .env:

• HF_TOKEN. Make sure that you got access to the diarization model
• OPENAI_API_KEY.

4. Run via Python:

# import all needed modules
from automatic_zoom_reports.asr.transcriber import Transcriber
from automatic_zoom_reports.asr.transcription import Transcription
from automatic_zoom_reports.summarization.summarizer import Summarizer
from automatic_zoom_reports.summarization.summary import Summary

# set your audio path
audio_input_path = "your_audio.mp3"

# initialize the process
transcriber = Transcriber()
transcription = transcriber.transcribe(audio_input_path)

# save transcript in one of the formats
transcription.save_html()
transcription.save_txt()
transcription.save_json()

# set token_usage_report_path to keep track of gpt token usage
token_usage_report_path = "token_usage.json"

# initialize the process
summary = Summarizer(token_usage_report_path).summarize(transcription)

# save summary in one of the formats
summary.save_html()
summary.save_txt()
summary.save_json()

5. Completed build and bot for connection to zoom is coming soon =)