Xiaomi open-sources MiDashengLM-Spatial audio model that locates sound sources

Xiaomi has open-sourced MiDashengLM-Spatial, an roughly 8-billion-parameter audio model under Apache 2.0 that describes sounds and reasons about their direction, distance and movement from binaural input.

The model builds on MiDashengLM-7B and adds a Spatial-Dasheng encoder, while mono input runs without directional information. Xiaomi says it will roll out the model gradually in smart home and car cockpit products, with no timeline given.