Automated Interview Transcribe and Proofread

This is a script to automate Interview transcription and proofreading and transcript formatting into .docx. Uses AWS Transcribe, S3, and OpenAI API (GPT3.5). Produces very high-quality transcriptions, even with very bad sound quality input files.

Installation

• Python 3.10+
• AWS IAM with access to S3, AWS Transcribe
• S3 bucket with public access - so this script can upload and read bucket contents.
• OpenAI API key
• See .envsample for what is needed

  pip3 install -r requirements.txt

• Rename .envsample to .env and save with your keys and bucket name.

Usage/Examples

usage: process_transcripts.py [-h] input_folder s3_folder

positional arguments:
  input_folder  Input folder with .mp4 videos
  s3_folder     Output folder name in S3 bucket

e.g:

python3 process_transcripts.py /Users/kvyb/Documents/Uitrial_Interviews testing_proofread

Features

• Uploads .mp4 2-speaker video to S3 bucket
• Transcribes speaker voices into text
• Proofreads the transcribed text, improving quality
• Formats the text into a .docx format. Output sample:

Feedback

If you have any feedback, please reach out to me.

Note:

• Only supports .mp4
• Still needs a quick manual proof-read. AWS Transcribe isn't perfect.
• Costs approximately $0.08 in total cost per 60-minute interview.