Motion matching
An animation technique that each frame searches a large database of recorded motion and plays the frame whose pose and upcoming trajectory best match the character's current state and the player's intent.
Instead of hand-built state machines and blend trees, where every transition between walk, run and turn has to be authored, you feed in raw motion capture and let a search find the next step. The result is responsive and turns feel natural, at the cost of memory and search time.
Ubisoft popularised the method with For Honor (2016) — Simon Clavet's GDC talk spread widely. "Learned motion matching" (2020) compressed the heavy database with neural networks so the technique fit on weaker machines too.