CoolFace
Datasetpublic

Hanno-Labs/sounds-like-fingerprints

sounds-like fingerprints Album metadata paired with an LLM-generated "sounds like" description, for ~25.7k albums. No review text. The copyrighted review prose is deliberately stripped — this dataset contains only factual album metadata and our own LLM-generated sonic fingerprints. column meaning artist, album album identity genre coarse genre tag year_released, rating, reviewer, label, reviewed factual metadata album_art_url cover image URL source provenance… See the full description on the dataset page: https://huggingface.co/datasets/Hanno-Labs/sounds-like-fingerprints.

sourceHugging Facecc-by-4.0updated 3mo agoView on Hugging Face
0likes25downloads
Dataset Card

sounds-like fingerprints

Album metadata paired with an LLM-generated "sounds like" description, for ~25.7k albums.

No review text. The copyrighted review prose is deliberately stripped — this dataset contains only factual album metadata and our own LLM-generated sonic fingerprints.

columnmeaning
artist, albumalbum identity
genrecoarse genre tag
year_released, rating, reviewer, label, reviewedfactual metadata
album_art_urlcover image URL
sourceprovenance of the text the fingerprint was generated from (scrape / statworx / orig)
sounds_likean LLM-written sonic description — vocals + emotional tenor + instrumentation

The sounds_like field is the target text that defines the geometry for the model **Hanno-Labs/langset-sounds-like**: it describes how each album sounds (with vocal register and emotional tenor weighted alongside instrumentation), so a langset latent trained on it clusters by sound, not by review text.

Fingerprints were generated with Gemini 3.1 Flash Lite from full album reviews; the reviews themselves are not included in this dataset.

Load

python
from datasets import load_dataset
ds = load_dataset("Hanno-Labs/sounds-like-fingerprints", split="train")
print(ds[0]["artist"], "—", ds[0]["album"], "::", ds[0]["sounds_like"][:120])