Séance investigates AI as a technological summoning apparatus—where neural networks trained on ghost sounds transform ambient noise into spectral voices. Using RAVE audio models and 6-channel spatial routing, I construct a ritual installation where phantom utterances emerge from feedback loops between microphones and speakers. Both EVP research and AI training operate through the same mechanism: pattern recognition within noise, conjuring presences that exist neither in signal nor imagination but at their threshold. The work renders audible the pareidolia at the core of machine learning—futures hallucinated from the static of the past, voices that were never spoken perpetually waiting to be heard.