Genius-Society/hoyoMusic

Intro

The hoyoMusic model is an innovative music generation model developed based on the Tunesformer architecture. The model has been trained on the MiHoYo music second creation piano dataset and is specifically designed to generate music inspired by user-created content from the MiHoYo user community. Currently, the training data is limited to abc score slices extracted from the music of MiHoYo's games such as Genshin Impact and Star Rail. With the Tunesformer model, the hoyoMusic model aims to capture the essence of the unique musical styles in the HoYoverse, which represents a fusion of advanced technology and artistic creativity.

Demo

https://huggingface.co/spaces/Genius-Society/hoyoMusic

Usage

from huggingface_hub import snapshot_download
model_dir = snapshot_download("Genius-Society/hoyoMusic")

Maintenance

GIT_LFS_SKIP_SMUDGE=1 git clone [email protected]:Genius-Society/hoyoMusic
cd hoyoMusic

Dataset

https://huggingface.co/datasets/Genius-Society/hoyoMusic

Mirror

https://www.modelscope.cn/models/Genius-Society/hoyoMusic

Reference

[1] https://huggingface.co/Genius-Society/tunesformer