Every input mode feeds the same 11-dimension engine.
The most natural route.
Start a voice session →The same engine is available over HTTP with your own key.
Whichever mode you pick, the engine returns the same 11-dimension model — every dimension carrying its own evidence, a confidence score and a stated uncertainty.
Prosody, turn-taking and interruption behaviour, read from a live session.
Word choice, directness and how you frame decisions, read from text.
Visible cues you describe yourself — expression, pace, how you handle friction.
The shapes above are structural illustrations of how each mode feeds the model — they are not measurements. Measured results are published on the case page.