Vano04/laions-got-talent-enhanced-precomputed-en

LAION's Got Talent Enhanced Precomputed English This dataset contains the precomputed embeddings of the LAION's got talent enhanced dataset english split at 16kHz. The audio was preprocessed with TuKoResearch/AuriStream100M_RoPE_librilight and the text transcriptions were preprocessed with google/embeddinggemma-300m. @inproceedings{tuckute2025cochleartokens, title = {Representing Speech Through Autoregressive Prediction of Cochlear Tokens}, author = {Greta Tuckute and… See the full description on the dataset page: https://huggingface.co/datasets/Vano04/laions-got-talent-enhanced-precomputed-en.