ArtifactNet-Real

Real (human-created) music dataset for AI-generated music detection research.

Overview

Split	Songs	Segments	Size
Train	15,546	70,463	47GB
Val	1,943	8,075	5.4GB
Test	1,944	8,384	5.6GB
Total	19,433	86,922	~58GB

Format

• Audio: WAV, 44.1kHz, Stereo
• Segment Duration: 4 seconds
• Hop Size: 2 seconds (50% overlap)

Structure

ArtifactNet-Real/
├── README.md
├── metadata.json        # Song lists per split
├── train/{song_id}/seg_XXXX.wav
├── val/{song_id}/seg_XXXX.wav
└── test/{song_id}/seg_XXXX.wav

Usage

This dataset provides Real (human-created) music samples for training AI music detection models.

To generate Demucs residuals:

from demucs.pretrained import get_model
from demucs.apply import apply_model

model = get_model('htdemucs')
# residual = original - sum(sources)

License

CC-BY-NC 4.0 (Non-commercial use)

Citation

@dataset{artifactnet_real_2026,
  title={ArtifactNet-Real: Real Music Dataset for AI Detection},
  author={...},
  year={2026},
  url={...}
}

Related

• ArtifactNet: AI-generated music detection framework
• Paper: [TBD]