The problem
Recommendation engines optimize for what you've already listened to, which is great for finding more of the same and useless when your mood doesn't match your history.
Playlists generated from how you're feeling, not what you last played.
Recommendation engines optimize for what you've already listened to, which is great for finding more of the same and useless when your mood doesn't match your history.
Take mood as the explicit input instead of inferring it from listening data, and build playlists against Spotify's audio features — valence, energy, tempo — rather than genre similarity.
Asking people to name their mood is friction, and self-reported mood is a blunt instrument. Inferring it would have been smoother and less accurate; making it explicit at least means a wrong result is correctable.
A working generator that turns a stated mood into a Spotify playlist. Feelings as a feature, not a bug.
05 Built with