Xiaomi open-sources MiDashengLM-Spatial spatial audio model
Xiaomi announced the open-source release of MiDashengLM-Spatial, publishing the model weights and inference code under the Apache 2.0 license. Built on MiDashengLM-7B, the model keeps basic audio understanding while adding spatial perception and reasoning; Xiaomi calls it the industry's first open-source end-to-end spatial audio understanding model. Reports say related technology has already been deployed in more than 30 applications across Xiaomi's smart home and car cockpit scenarios, and the new model will be rolled out to those scenarios.
Spatial perception extends audio models from recognizing sound content to judging where sound comes from, and Xiaomi's release of weights and inference code gives developers a base for spatial audio research and application development.