Skip to content

paarthtandon/TrashGPT

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

94 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

TrashGPT

Welcome to our repository. This project was completed for CS 685 (Spring 2023). The goal of this project is to generate clips of the Trash Taste Podcast. Each folder specifies one aspect of the data pipeline.

Setup

There are many different subroutines in this repository, and I would recommend only installing the packages needed for that subroutine to avoid package conflicts.

  1. Install Python 3.10.X
  2. Setup a virtual environment: python -m venv .venv
  3. Activate virtual environment. Command is platform dependent.
  4. Install PyTorch. Make sure to use the command on their site. It is platform dependent.
  5. Install other requirements as seen in the README of the pipeline component you are running.

Pipeline

  1. Data download
    • Create directory raw_data/, move into the directory.
    • Run yt-dlp.exe --yes-playlist --write-sub --write-auto-sub --sub-lang "en.*" --sub-format json3 -f m4a https://www.youtube.com/playlist?list=PLUHmmIt9sU6i4JlDABqLeWybD_ZrJf9LB
  2. Data preprocessing (look in data_prep/)
    • Speech diarization
    • Speaker naming tool
    • Dataset formulation
  3. Training the models and generating transcripts (look in modeling/)
    • Training and generation code for GPT2-Small, GPT2-Med, Bloom 560M, and LLaMA 7B
  4. Voice cloning and audio performance generation (look in voice_cloning/)
    • Transcript parser
    • TTS with voice cloning using tortoise-tts

Results

Other Documents

About

Automatic podcast generation.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 3

  •  
  •  
  •