To get a transcript from a Craig recording, first, use the Craig bot to record your Discord voice channel to get separate audio files for each speaker.Then, use a separate transcription tool like TASMAS (for automatic transcription and summary), JotMe or another AI-powered transcription service to convert the audio files into text. 


Record with Craig Bot 

  1. Invite Craig: to your Discord server from the Craig website.
  2. Join the voice channel: you want to record by typing /join in the chat.
  3. Stop the recording: by typing /stop.
  4. Download the audio files: from the link you receive after the recording ends. (Download Ogg format)


Transcribe the Audio Files

Tasmas is an open-source project that takes your separate audio files and generates a single, coherent, attributed transcript.

You can find it on GitHub. 


Run from Docker (Didn't work on my MAC)

docker run -it -v /Users/john/Desktop/recording:/recording --gpus all kaddaok/tasmas:latest

On my Mac, I had to remove the --gpus all flag.

docker run -it -v /Users/john/Desktop/recording:/recording kaddaok/tasmas:latest


To run TASMAS, you must provide at minimum:

  1. an operation mode (recognize, assemble, summarize, semiauto which does the first 2, or fullauto which does all 3) and
  2. a folder path to process.


tasmas semiauto /recording/test


Building Locally (Code Changes Required!)

Prerequisites

 brew install portaudio


Clone and Build

git clone https://github.com/KaddaOK/TASMAS.git
cd TASMAS
pip3 install whisper-timestamped
/Applications/Python\ 3.13/Install\ Certificates.command
export SSL_CERT_FILE="$(python3 -c 'import certifi; print(certifi.where())')"
pip3 install .


Run

python3 tasmas.py semiauto /Users/john/Desktop/recording/20250901




  • No labels

Copyright JMEHAN.COM