🔊 MiDashengLM-Spatial
Unifying General Audio Understanding and Spatial Awareness
The first open-source end-to-end unified audio-language model that handles both general audio understanding and spatial (binaural) awareness in one architecture. Upload stereo / binaural audio, then ask anything — including where the sounds are coming from.
Answer
Try an example (bundled binaural clip from the model repo)
Notes
- Mono audio still runs, but carries no directional information — use stereo recordings for spatial questions.
- Audio is resampled to 16 kHz and capped at 2 minutes.