MLAAD-tiny is a very small subset of the full MLAAD dataset, designed for education, prototyping, and debugging.
Many teaching environments (e.g. Colab, Kaggle, university notebooks -- se this notebook for example) impose strict storage limits, which makes large-scale audio deepfake datasets impractical to use. To address this, we provide MLAAD-tiny, a compact yet representative version of MLAAD.
git lfs install
git clone… See the full description on the dataset page:
https://huggingface.co/datasets/mueller91/MLAAD-tiny.
