Speaker recognition with two-step multi-modal deep cleansing

Connections are not prepared for this record yet, or its source metadata has changed. The original record remains available while background snapshots are rebuilt.