albinati / audioSummaryGPT

This project basically uses Whisper to transcribe an audio file and GPT models to summarize the content.

Audio Summary Generator

This project uses OpenAI's GPT model and the Whisper ASR system to generate a summary from an audio file.

Setup

1. Clone the repository

   git clone https://github.com/albinati/audioSummaryGPT.git
   cd audioSummaryGPT
   

2. Install the required Python libraries

   pip install -r requirements.txt
   

3. Set up your environment variables

   Create a .env file in the root of your project and fill it with your own values:

   # .env file    
   OPENAI_KEY=[your-key]
   INPUT_FOLDER=./content
   AUDIO_FILENAME=file.wav
   OUTPUT_FOLDER=./output
   OUTPUT_FILENAME=summary
   TRANSCRIPT_FILENAME=transcript
   CHUNK_FILENAME=chunk
   USER_PROMPT1=Faça um resumo do arquivo em bullet points 
   USER_PROMPT2=Corrija possiveis erros na transcrição
   AI_ROLE=Você é um assistente expert em resumo de textos 
   GPT_MODEL=gpt-3.5-turbo
   GPT_TEMP=0.7 #Default Temperature

Usage

1. Transcribe the audio

   Run the transcriber script to convert your audio into text:

   python transcriber.py

   This will create a transcript.txt file in the output folder.

2. Generate the summary

   Run the summarizer script to generate a summary of the transcript:

   python summarizer.py

   This will create a summary.txt and summary_summarized.txt files in the output folder with the summary.

Notes

The transcription and summary quality can be affected by the clarity of the audio and the specific GPT model you're using. The Whisper model is used for transcription, and GPT-3.5-Turbo or GPT-4 can be used for summary generation. The .env file should contain all the necessary configuration.