Currently GDevelop does not preload audio, so you have to play the sound at the beginning of the game for at least a few seconds to preload it manually.
The way I’m thinking about addressing this in the game I’m making is to do a fade-in (black box covering the entire screen, increasing it’s transparency using a TimeDelta() expression until it’s invisible. Potentially a very fast one.
I’ll then play all of the music and sounds for my level at 0 volume to load them into memory, and start the music over at normal volume once the transparency is 0. That should reduce or eliminate the perceived delay, maybe?