Some audio blocks should take a sound object as input

[ToonTalk]

Speech recognition won't work but passing it to a model should.