Single-photo mode already computes most ratio metrics steadily — so what does camera precision mode add? In one sentence: it swaps 'a photo of unknown provenance' for 'a live capture under controlled conditions'. It opens the front camera right in the browser page (nothing to install), acts like a photography director correcting your conditions in real time, then hardens the measurement with multi-frame data.
Live guidance: three virtual directors
- Distance guidance. The iris's size in frame yields a live estimate of camera distance; too close, and the prompt says 'pull back to arm's length or more'. This kills perspective distortion — selfie range can inflate the nose-width ratio by roughly a third, the single biggest error source of photo mode.
- Pose guidance. Head yaw and pitch are monitored continuously, with prompts to square up — ensuring angle-sensitive metrics like the Three Courts and symmetry are captured truly frontal and level.
- Stability gating. Once every condition passes, the tool captures a burst of frames and averages each landmark's position across them — single-frame detection always jitters at the pixel level, and multi-frame averaging visibly crushes that noise.
Iris calibration: an invisible ruler inside the photo
Pure ratios never need the face's real size — but judgements like 'a broad frame versus a fine one' do. Precision mode borrows a physiological constant: the adult iris varies remarkably little in diameter across people, making it a natural reference object present in every shot. Calibrating against it yields a rough estimate of real dimensions such as cheekbone width, with baselines adjusted for different populations. Honesty required: this is approximate-grade estimation, sensitive to conditions, and the report labels it as such — it extends the dimensions measured, and never impersonates millimetre precision.
Which fine metrics it unlocks
With capture quality raised, four families of detail become measurable that never were before: brow detail (length and density — the true substance of the Siblings Palace), silkworm contour (the Children Palace's millimetre-scale tissue), nose-root reading (the Resilience Palace, optionally aided by a guided profile shot for contour), and complexion assessment (the Fortune Palace — brightness and evenness only compare meaningfully once light is controlled). This is exactly why several palaces in the basic report carry a 'pending precise measurement' tag: nothing is being withheld — a single photo genuinely cannot hold those metrics steady, and precision mode supplies the missing conditions.
One use fits the tradition's own logic: periodic retesting. Bone-structure ratios hold for years, but complexion moves by the month — the old books themselves file it under 'changeable'. A monthly recapture under fixed conditions, read as a curve rather than a point, is precision mode's most valuable long-term habit.
Finally, the technical boundary: precision mode is not a 3D scan. Reconstructing a metric 3D face from one photo carries a scale ambiguity in principle, and published benchmarks show most methods add little individual accuracy — so we chose the humbler route: controlled capture, multi-frame averaging, iris calibration. Honest measurement begins with not overselling itself.