That actually seems to set the language for it to transcribe (as opposed to it guessing), with the following triggering a translation to English:

result = model.transcribe("audio.wav", task="translate")

But your post helped me figure out the above, so thank you!