
Solo AI tunes a music model on listening histories people contributed.
Fine-tuning a generative music model on listening histories that a community chose to share.
The data
Skips, replays, listening sequences and time-of-day patterns from thousands of people.
OpenDataLabs collaborated with Solo AI to build the dataset for fine-tuning its generative music model. Catalogue metadata describes what a track is. This data shows how people respond to it.
What people agreed to
Members of a community interested in AI-generated music chose to contribute their listening history.
The record
Every record was consented to at the source and traces to a person who chose to take part.
What it made possible
Solo AI fine-tuned its generative music model on real preference signal, beyond what catalogue data provides.
Solo AI used the dataset to seed reinforcement learning in fine-tuning. Its model now generates music that reflects real human listening behaviour.