Xiaomi open-sources audio AI that can locate the sounds it hears
Xiaomi has released MiDashengLM-Spatial, an audio model that can identify sounds and reason about where they come from. The company published its model weights and inference code under the Apache 2.0 licence, and says it plans to introduce the technology gradually in smart-home and car-cockpit products.
Bottom line — Xiaomi says the new model will move into smart-home and car-cockpit products, but has given no timeline or named products.
Go deeper 4
-
TechNode reports the model retains speech recognition, audio captioning and questions about environmental sounds and music.
-
According to zglg.work, directional cues require binaural, two-channel audio; mono input does not provide direction.
-
Xiaomi describes it as the industry's first open-source end-to-end spatial audio model; zglg.work notes that this claim comes from the company and its researchers, not independent assessment.
-
Xiaomi says earlier Dasheng-series models are already used in more than 30 smart-home and automotive applications; that figure does not refer to MiDashengLM-Spatial.