Community · E2 · artifact verified
Drive a 3D face with typed emotion judgments
A 3D face that reacts in real time to what you type: while you write, Jev judges which emotion the recipient would feel and how strongly, and the calibrated answers drive 52 ARKit blendshapes on a face mesh.
01 · Role in the system
What Jev does here
The demo inverts emotion AI: instead of reading a face, it predicts the face on the other side of the message. Each keystroke sends the draft as typed questions - which emotion would the recipient feel, how strongly - and the calibrated answers drive 52 ARKit blendshapes in real time on a 3D mesh. The judge answers about the recipient's feeling, not the text's sentiment, and the mesh renders the probability-weighted intensity. Deployed as a Vercel-hosted API route plus web client; the judgment path is a single evaluate.js module.
02 · Control boundary
Where Jev sits
Jev as the recipient-emotion judge of a live typing surface: the draft becomes typed emotion questions per keystroke, calibrated answers drive blendshape intensity on a 3D face, rendering the felt impact of what you write.
Code owns the loop, permissions, thresholds, validation, and side effects. Jev owns only the bounded judgments described above.
03 · Known limits
What this evidence does not prove
- Emotion labels are recipient-modeling approximations; no user-study validation.
- 0 stars; created 2026-10-11, single-maintainer demo.
- Blendshape intensity mapping is hand-tuned.
04 · Attribution
Public sources
This is a Community record: the project was published by a third-party community author.
- Imjungjuna ↗Community · github · public · checked 2026-10-12