This is an a POC for a complementary tool used by ARTAX.
The goal is to investigate the usability of generated lyrics used for sourcing algorithmic music concepts.
Client access token generated via https://genius.com/api-clients and kept in a plain textfile called ´genius.txt´
Model training should be performed in Colab notebooks with GPU.
- Scraper implemented using Genius API to pull all lyrics by artist
- Basic data cleaning
- Model building pipeline implemented
- Generation and note mapping tested with wovels and phonetics
- Forther data cleaning options (repetitions, etc. )
- Investigate lyrics filtering before moving to corpus based on [semantic classification, Flesch Kincaid Readability Test score, ]
- Incorporating TL Models, and Style match architecture