Fucked Up and Bad AI

Machine learning models are a purely product of their dataset. It is unethical to train generative models on other's work without their consent. While I will try to stay true to that principle throughout this project, I will at times use permission as consent and use public domain datasets (like LibriSpeech) instead of volunteer or paid datasets. If you are in one of these datasets and do not like it contact me and I will remove you from my copy. I will never knowingly use works which remain copyrighted or work from people who refuse to be used in datasets.

LyingBard

• LibriVox through LibriSpeech
• (Pretrained) Simple Speaker Embedding