A voice a family can still reach

An interactive digital persona is a consented, evidence-backed likeness of a family member — their voice, appearance, and remembered detail — that relatives can talk with inside their private Family Connect site.

A persona is built while the person is able to take part, from their own recordings, photographs, written answers, and the memories relatives contribute about them. Audio personas are included free; a lifelike video persona is a one-time upgrade.

The member records their voice, frames a portrait, and answers interview prompts about their life, work, humour, and beliefs. Relatives can add what they remember, and each contribution is stored as evidence with the contributor's name attached.

Nothing is published automatically. A steward reviews contributed evidence, and the subject sets rules about what the persona may discuss and who may speak with it.

The Forever Spoken Trusted Multi-layered Fidelity Index is the scoring model that measures how well a persona is evidenced across five layers: identity, personality, provenance, uncertainty, and governance.

Each piece of evidence is weighted by who contributed it, how close they were to the subject, and which period of the subject's life they witnessed. A persona shows its score openly, so a family always knows how well-grounded it is rather than how convincing it sounds.

Access is decided by kinship and by the subject's own rules. Seven access tiers run from immediate family through to distant relatives, and the subject or their steward can tighten access at any time.

Platform-level governance sets a baseline every family inherits. A family site may tighten those rules further, and an individual member may tighten them further still. No layer can loosen what the layer above it set.

A deepfake imitates someone without their knowledge. A digital persona exists only with the subject's recorded consent, and the subject controls what it may say.

A persona also refuses to guess. Where the evidence is thin, it says the record does not hold that, rather than inventing an answer in the person's voice.