The Principle
A voice clone captures a moment. A LifeLine captures the arc.
A person's voice is different at different times and contexts of life — at 25 vs 55, excited vs grieving, telling a story vs reading to a child. The Voice Library is not a single recording. It is a living, continuously updated record of a human voice across a life.
It should be constantly referenced. Always updating. Never static.
What It Captures
- Age-indexed samples — voice across different decades when available
- Context-tagged recordings — storytelling, instructional, emotional, celebratory, intimate, public
- Continuous updates — every voice note, audio message, video adds to the model
- Historical recovery — old voicemails, home videos, recordings pulled from wherever they exist
How It Grows
Every time a user:
- Sends a voice note
- Records a LifeLine memory
- Answers a guided prompt
- Shares a video
- Uses Wayfinder narration
...the Voice Library grows. Passively. Continuously. Without effort.
What It Enables
The Ripple (virtual persona) speaks from the Voice Library — not from one recording but from the full range of a person:
- Johnny at 30, excited about a new idea
- Johnny at 55, thoughtful and reflective
- Johnny reading a story to a grandchild
- Johnny explaining something he cares deeply about
The grandchildren choose which Johnny they want to be with.
The Living Record vs. The Portrait
A portrait is a moment. A living record is a person.
The Voice Library is the difference between a recording and a presence. Between a headstone with dates and a rich tapestry.
Technical Notes
- ElevenLabs Voice Cloning as the foundation
- Multiple clone versions indexed by age/context
- Continuous fine-tuning as new samples arrive
- Privacy-first — voice data stored on decentralized infrastructure, owned by the individual
- Explicit consent required for Ripple voice generation
- Separate consent tiers: family access vs. researcher access vs. public
For Johnny's LifeLine
- [ ] Initial voice recording session — multiple contexts
- [ ] Recovery of existing voice material (videos, voicemails)
- [ ] Ongoing: every voice note tagged and added
- [ ] First clone created when sufficient material exists