J
JARVIS OS
Reflect Mirror
Avatars

Faces coming into focus.

Every avatar training across the GPU fleet, ranked by mouth precision — the metric that decides whether the smile in the mirror is real enough. 3 of 3 machines are working right now.

BEST MOUTH22.5dBml-0222.5dBcart-0216.1dBcart-0120.2dB3 of 3 TRAININGcolor = mouth PSNR · pulse = active
Leading the fleet
exp_opt_E3b_ml02_tuned_a
on ml-02 · step 30K
22.5
dB
Photoreal
face
27.6
identity
0.65
splats
100K
step time
193ms

Mouth Leaderboard

sorted by mouth precision · the metric that decides everything
exp_opt_E3b_ml02_tuned_a
📡 ml-02·step 30K·100K splats
mouth
22.5dB
Photoreal
trajectory
face 27.6
csim 0.65
2
exp_opt_E3b_cart01_grant_baseline
📡 cart-01·step 30K·100K splats
mouth
20.2dB
Strong
trajectory
face 24.7
csim 0.73
3
exp_252_pro_validation_grant
📡 cart-02·step 33K·100K splats
mouth
16.1dB
Promising
trajectory
face 16.4
csim 0.04
Want the deep view?
The full ML training dashboard with visual checkpoints, gaussian splat viewer, and mirror metric arc lives next door.
Open ReflectIris ML →
Chat with JARVIS
Connecting...
Telegram
💬
No messages yet
Just talk to JARVIS like a person. No commands needed — ask anything about the project.
Same conversation as your Telegram chat. JARVIS thinks before replying.