I decided to try out the new audio training option in Training Studio. I have no idea what I'm doing. Given I had to run FFMPEG from command line to make a sample "video" (the image in gallery has audio. Unmute it) to even post this, and correct for an inconsistently named base model I don't think Civitai does either! The ongoing Asuka Apocalypse likely already told you that.

Trigger: A Golden Sun style battle theme for

Trained on the 15 battle themes from the original duology. Autocaptioned with manual edits to replace (incorrect) references to chiptune and video game generation (which appear to have been assigned at random) with the trigger, plus a few very inaccurate time codes were removed. Also imported model needed its base model set manually, presumably because it tries to set base model to "ACE-Step" (as trainer lists it), but fails because there's no such category, only "ACE-Audio".