Astra-CLR works with photometry data — records of how a star's brightness changes. Like a listener who learns to tell jazz from rock without a textbook, the program slices light signals into fragments and searches for repeating rhythms. It brings pieces of the same "stellar melody" closer together and pushes alien ones apart, gradually building a musical map of the sky. Just a few labeled examples are enough to sharply boost its recognition accuracy.

This "ear" for stellar rhythms is especially valuable for future surveys that will count objects in the billions. Precise light analysis ( spectroscopy ) is only possible for a tiny fraction of objects, while Astra-CLR quickly sorts all the "music", helping to study stellar evolution and the nature of supernovae. This method comes from speech recognition systems: the algorithm learns just like a baby picks out words from a stream of sound, without yet knowing their meaning.