Introducing MichiAI

MichiAI is a low-latency, full-duplex speech model. No transcripts, no delay, just pure end-to-end speech generation on consumer GPUs.

Part 1: Speech Generation