The music bed is the masking audio: a track that loops for the whole session and masks the voices at the level you set. In the Music phase, clicking a track auditions it; nothing joins your program until you press “Use in project”, so you can A/B candidates freely.
Browse by mood opens the library: one click narrows the whole collection to a feeling, running from calm and dreamy through uplifting, romantic and nostalgic to melancholic, dark and epic, with meditation, sleep and focus at the end for when you know what the session is for. Clicking another mood switches. The chips are the same tag filters the search and filter controls use, so the full tag list is still there when you want a finer cut.
Once your voices are generated, Hear with your voices on the audition deck plays a candidate under them through the main player. It is clearly labeled as a trial and not saved until you use it. While the trial runs, clicking another track swaps the music under your voices in place.
Beds are loudness-matched when they land to reduce volume jumps while switching. A bed that swings from loud to quiet can be smoothed with Even it out. The studio measures each new bed and turns it on by itself when the track plays unevenly. Flip the switch yourself once and your choice sticks.