suno_audio/README.md

metadata

license: mit
task_categories:
  - text-to-audio
  - audio-classification
tags:
  - music
  - audio
  - suno
  - music-generation
  - ai-music
pretty_name: Suno Audio Dataset
viewer: true
size_categories:
  - 10K<n<100K

🎵 Suno Audio Dataset

A comprehensive dataset of 49,698 AI-generated music tracks from Suno, organized in 50 batches of 1000 samples each.

🎧 All audio files are playable directly in the dataset viewer!

Dataset Structure

The dataset is organized into batches (batch_0, batch_1, etc.), each containing up to 1000 audio samples with metadata.

Fields

• audio: 🎵 Playable MP3 audio file (click to play in viewer!)
• id: Unique track identifier
• title: Song title
• display_name: Creator/artist name
• handle: Creator handle
• tags: Music tags, genres, and styles
• prompt: Text prompt used for generation
• duration: Track duration in seconds
• play_count: Number of plays on Suno
• upvote_count: Community upvotes
• model_name: Suno model version used
• created_at: Creation timestamp
• status: Track status
• is_public: Public visibility flag

Usage

Load Entire Dataset

from datasets import load_dataset

# Load all batches
dataset = load_dataset("Humair332/suno-audio")
print(f"Total tracks: {len(dataset['train'])}")

Load Specific Batch

# Load only batch 0
dataset = load_dataset("Humair332/suno-audio", data_dir="batch_0")

Play Audio

# Get audio data
audio_data = dataset['train'][0]['audio']
audio_array = audio_data['array']
sampling_rate = audio_data['sampling_rate']

# Play in Jupyter/Colab
from IPython.display import Audio
Audio(audio_array, rate=sampling_rate)

Filter by Tags

# Filter by genre
rock_songs = dataset['train'].filter(lambda x: 'rock' in x['tags'].lower())
print(f"Found {len(rock_songs)} rock songs")

Most Popular Tracks

# Sort by play count
from datasets import Dataset
df = dataset['train'].to_pandas()
top_tracks = df.nlargest(10, 'play_count')[['title', 'display_name', 'play_count']]
print(top_tracks)

Dataset Statistics

• Total Tracks: 49,698
• Batches: 50
• Batch Size: 1000
• Format: Apache Arrow with embedded MP3 audio
• Audio Format: MP3
• Metadata: Tags, prompts, engagement metrics

Explore the Music! 🎶

Click on the dataset viewer above and browse through the tracks. Click any row to play the audio directly in your browser!

Source

Original dataset: nyuuzyou/suno

License

MIT License

Citation

If you use this dataset, please cite:

@dataset{suno_audio_dataset,
  title={Suno Audio Dataset},
  author={Humair332},
  year={2026},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/Humair332/suno-audio}
}