The measured warp-flash shape.
A rounded hump, and it starts before its start frame. The reference's
f1085 and f2382 are threshold crossings in the shot detector, not the frame
the light begins — exactly the distinction THRESHOLD_FRACTION draws for the
titles, met again here. Tracked whole-frame means for the first flash, as a
fraction of its own peak, run 0.36 at t=0 and 0.56 at t=1: a third of the way
up before the envelope had left zero, which a capture reads as a flash two
frames late. So the ramp opens at t = −3.5.
And the top is rounder. The old envelope held exactly 1 from t=4 to t=11
while the reference rises 0.89 → 1.00 → 0.83 across those same eight frames —
a plateau a host cannot shape, because a constant input carries no
information for it to shape. Peak sits at t≈6–8, and smooth saturates, so
this pair still holds exactly 1 across t ∈ [5.5, 9]: three and a half frames
rather than eight, not none. Shaping those last frames needs a curve that is
not a min of two clamped ramps, and that is a re-meter against the reference
rather than an edit.
One function for both flashes still: their normalized shapes agree. Their amplitudes do not — the second peaks 17% brighter in output mean — which is a fact about the two shots, not about the envelope, and is left to the script.
- startFramenumber
- Undocumented
- framenumber
- Undocumented