Do you need a real audience for the Feynman technique — and what does NOT count as doing it?

No real audience is needed — a rubber duck, pet, or imaginary student works — but you must do it FROM MEMORY, and just re-summarizing in the original jargon is NOT the technique.

The mechanism is retrieval plus plain-language generation, so what matters is that you pull the explanation out of your own head and translate it into simple words. Talking to a rubber duck, a pet, or an empty room is fine — the "audience" only forces the plain-language framing. Two things break it: reading from your notes (that's not retrieval), and re-stating the concept in the same technical vocabulary (that skips the translation step where gaps reveal themselves). If your "explanation" still leans on jargon, you haven't tested your understanding.