By pressing Ctrl+Delete. Audio-based cards are completely useless.
Unlike text-based sentence cards, audio-based cards are easier to cheat on. You end up memorizing the answer based on the audio clip which doesn't necessarily help you remember the target word.
Doing audio-based cards has no effect on kanji ability. To learn how to read kanji one needs to train reading kanji.
It is also unclear whether audio-based cards can actually help develop listening ability because listening to the same audio clip over and over is not the same as listening to different representations of words being spoken in authentic content.