Android app, where a group can log in, and start a multichannel voice recording session with server side blind source separation. Phones should be placed scattered among the recording scene, and blind source separation should be able to create distinct channels of voice sources.
Synchronization of the devices would be explicitly acoustic in the first implementation, automatic later.
Should be useful for understandable records of simultaneously speaking group (meetings, boards), acoustic music groups.
Data science part: blind source separation is purely statistical approach, any non-gaussian source should be possible to be identified and amplified.