Become a capsule
step 1 of 7
  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
Patent #19/319,614 · live demonstration

Become your own capsule.
Talk to yourself, signed.

In about five minutes you will create a working Personality Vector of yourself — your voice, your face, ten of your formative memories, and your cognitive style — and have a conversation with the resulting capsule. Every reply will be cryptographically signed and verifiable.

1Face

Five short webcam frames — neutral, slight left, slight right, gentle smile, mild surprise. Used to build a photo avatar.

2Voice

Read one short paragraph for ~30 seconds. Used to build a voice clone that speaks your replies.

3Memories

Twelve structured questions about formative moments, values, advice you would give. Used to build your Personality Vector.

4Talk to yourself

Your face, your voice, your memories, your seal. Same architecture that runs Will Carrington above.

Privacy by design — patent claim 11

Your raw streams never leave your device.

On-device feature extraction

Webcam frames are JPEG-encoded in your browser before upload. The video stream itself is never recorded — we only retain five still frames you choose to send.

Prosodic feature extraction

Voice is captured as a single short clip. We send it once for cloning, then both we and the cloning provider discard the raw audio after the voice profile is built.

One-hour TTL

The temporary capsule we build for you exists in memory for sixty minutes after your last interaction. After that, the avatar id, voice id, and capsule are auto-deleted.

No model training

Your data is never used to train any model. The Personality Vector is a structured object stored alongside our seal key — it is not gradient signal.

You can quit at any step. Nothing is sent until you click a confirm button.