datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
acestepsongsacestepacestep-v15-turbo-synthetic-60sace-step-multilingual
ACE-Step Multilingual Prompt Language Dataset
Dataset accompanying:
"Does Prompt Language Affect AI-Generated Music?
A Multilingual Evaluation of ACE-Step 1.5 Turbo"
Dataset description
250 instrumental tracks generated using ACE-Step 1.5 Turbo.
5 prompt languages
10 musical prompt families
5 matched random seeds
30 seconds per track
48 kHz
identical generation settings across language conditions
Languages
English
Swedish
Spanish
German
French… See the full description on the dataset page: https://huggingface.co/datasets/jacobrrak/ace-step-multilingual.ACEStep-Songssongs generated by ACE-Step
score_lyrics field represnets the scores given by gpt-4o for the lyrics(1-10, only those >=8 are preserved), -1 for instrumental piece
the full lyrics-tags-score dataset is at https://huggingface.co/datasets/Yi3852/lyrics-tags_gen
more info see https://github.com/ace-step/ACE-Step/issues/313
Citation
@misc{jiang2025advancingfoundationmodelmusic,
title={Advancing the Foundation Model for Music Understanding},
author={Yi Jiang and Wei Wang… See the full description on the dataset page: https://huggingface.co/datasets/Yi3852/ACEStep-Songs.acestep-vocals-lora-demo-audioACE-Step-1.5.7zACE-stepace-step-finetune