Voice Lock starts automatically in supported buffers. It maps faces and structural text properties to personalities, making comments, strings, headings, links, and other text sound distinct without extra words.
Compare two styled elements in code, Org, Info, or EWW. Toggle C-e d v locally, read the same elements, then toggle again to restore. C-e d C-v changes Voice Lock globally. Disabling it removes that layer of voices, not ordinary speech, semantic feedback, tones, or icons.
The result depends on engine and presentation settings. Voice Lock, explicit face rules, and semantic rules are separate; see Voice Lock in Aural Presentation. Reusable voice, sound, and profile choices belong to that chapter rather than to a Voice Lock toggle.