A voice was, until recently, recognisable only to people who had heard it. A recording of a voice could be matched to a known speaker by someone familiar with them, and otherwise its association with a particular person depended on context: a name, a setting, an introduction.

This is no longer reliable. A recorded voice is now, like a face, something that can be searched. A sample of speech can be tested against other recordings in a way that establishes, with high confidence, whether the same person was speaking. The technology is general-purpose, widely available, and does not require any prior knowledge of the speaker.

The consequence is that voices, like faces, have become connectors. A recording of a person speaking at a public event, a podcast appearance, a corporate video, an interview clip, a voicemail left in another context: each, taken separately, was anonymous to anyone who did not already know the speaker. Together, they are a single thread.

Older recordings carry particular weight here. A voice given to a microphone twenty years ago, in a context the principal has long forgotten, becomes connectable to their voice today. A recorded conversation that was, at the time, of no consequence, becomes part of an assembled picture that the principal has had no opportunity to consider.

The work in this category, for a principal whose voice has been in circulation for a long time, is patient rather than dramatic. It begins with understanding what recordings exist, where they sit, and how they connect. The recordings cannot, in most cases, be unwound. The connections they are part of can, with care, be made less easily assembled.