constants.module
Versions
Every version of this package its world published, newest first by when it first appeared. A version published again later keeps its place.
verify: shot 12 is printed against the room the fall leaves, and the mix keeps three decibels of headroom
01M2DGMTKWDY96343525TN3AR4d9f214bbffBrowse filesCompare with previousCompare from hereverify: the 85 mm inserts resolve, shot 13 clears the wreck, shot 12 is printed for the room the fall left, and the mix keeps its headroom
01M2DFDB6NQ6KZMHSC8P5ERC09051adfbd6dBrowse filesCompare with previousCompare from herelook: the print solved against the room, and the red kept out of the pink The exposure response of a frame to a stop of trim is not a constant - it runs from 1.2x for a frame already on the shoulder to 8x for one in the toe - so each shot's trim was solved from its own measured response rather than from a slope. Over shots 1 to 13 the delivered medians now run 47 to 70 (a spread of 1.49, against 2.86 before), no shot clips a pixel, and the two cuts that jumped two stops - shot 3 to 4 and shot 8 to 9 - are flat. - The fall's light loss is taken back off INSIDE shot 11 and handed over again by shot 12, so the frame after the impact is darker than the frame before it, which is the reading of the beat. - Shot 8's 85 mm cut-in on the glass carries its own trim for the second and a half it is on screen: a dark subject at close range inside a shot printed for the room behind it. - The grade's chroma gate opens earlier and harder, and a hit that arrives with more blue in it than green has that blue taken back out - cadmium has none. The curtain's dominant mode goes from R/G 2.00 to 2.38 and the red's share of the frame from 29% to 9%, which is the share the film allows it. - The tone curve keeps a highlight's hue almost all the way up: this film has three saturated sources and each is meant to hold its colour into its core. - reinstall() drops the grade cube, so an edit to the grade reaches the frame.
01M2DBZTR2Z2S9KHWNDR747P5B99043d1d71Browse filesCompare with previousCompare from hereair: the storm runs on the film clock, and the rain is rain again AirClock reads cutscene.status().time while the film plays instead of gating on play mode, so all four strikes evaluate and the fog, shafts, rain rates, water column and wetness ramp move continuously rather than stepping at the fourteen shot beats. Three samples of AirClock:filmTime() beside cutscene.status().time now agree to 0.000 s (they were 21.0 against 37.2). The shot-1 strike lights the gate: 11.70 light intensity at t=4.50 against 0.0 at every second before. AirClock and MothSwarm tick once a frame off engine.mode rather than off an os.clock heuristic; no wall clock is read in the slice. The twelve particle emitters are one EMITTERS table, and demo.air.reset() re-asserts it on the live scene. A built scene restores its entities from the scene record, and a ParticleEmitter's record does not carry the nested spec keys the component documents, so every emitter in the loaded world was a POINT emitter at schema defaults — which is why nine thousand rain drops fell from one spot and read as a white pipe over the facade rather than as rain (#12889). Exteriors: the height fog runs at a tenth of the interior density until t=14, which takes the hard arc out of the sky over the gate and gives shots 1 and 2 real blacks again. A near-lens rain sheet carries the storm at reading size. The volumetric raymarch is 24 steps / 4 shadow taps over 70 m, with density and lightScale rebalanced: vol_scatter median 5.774 ms at 77% attributed, against 17.041 ms. The plaster and glass bursts fire on b11_impact, when the chandelier lands, rather than when the cable parts. The four empty bNN_flash handlers are gone; their cues are still declared here and stagecraft hangs the thunder on them. READMEs for AirClock, MothSwarm and glow.shader.
01M2DB6VW8JTGV0DV3RKDTGVERe796698962Browse filesCompare with previousCompare from herehall: the room's mechanism answers its callers and starts at rest Every method another file calls is now declared on public: with a colon and called with one, so the reform, the lids, the glow and the chew stop raising on every drive: HallFaceRig:driveTo, HallEye:openTo/glowTo, HallFloorChew:chewTo/ setGrin, HallBreath:rewind, HallDrift:rewind, HallRun:recapture, and the six call sites in hall.module. demo.hall.reset() reached its first statement for the first time; beats.reset() now reports five slices instead of four. Every driven field is authored at its shot-1 value on component.add - value, target and rate - so a scene record saved mid-film cannot open the film in a room that has already become the face. M.restPose() writes the same six through the rig's own methods and is reset()'s first statement. The schedule of the room is the stagecraft slice's FaceRig, which holds all four on continuous curves across the whole film. This slice's beat handlers stop driving them; what is left on a beat is the lit window and b02_windowout. Eight components route update and editorUpdate through one engine.mode guard. The film runs in edit with the simulation going, where the engine fires both, so HallBreath's 8.6 s breath was a 4.3 s breath, HallMirrorLag's 0.34 s lag was 0.17, and HallFloorChew wrote 176 tile transforms twice a frame. Three reflection probes, one per room the film stands a reflective surface in, so the mirror corridor and the wet checkerboard reflect the rooms rather than the sky; M.bakeReflections() re-bakes them from the light standing now. The marbles come down to roughness 0.12/0.18 so the floor answers as wet stone. The five primitive meshes are resolved once through a static table like the materials already were, the two dead K.L locals and their silencers are gone, and MOUTH_WIDEN joins the other five lip constants so buildMouth and HallFaceRig read one set of numbers. Nine READMEs written; no scaffold left in demo/hall.
01M2DAET5FEV8W2945MF55BMK8729e917d9fBrowse filesCompare with previousCompare from herestaging: the shot table restaged, and the two moments every slice has to agree on K.IMPACT (70.75) is the second the chandelier lands and K.DRESSING carries the bulb run, so the sound, the light drop, the plaster burst and the glass all read one number instead of four. Shot 8's 85 mm cut-in moves onto the shot row as K.SHOTS[8].insert and film.module reads it, along with every duration, from the shot table — and refuses to load if the eighteen camera moves no longer sum to the film's hundred seconds. Seven shots are restaged to what the spec's shot list now reads: shot 1 stands back so the whole grin of the arch is in frame with the ground under it, shot 3 stops short of the threshold so the ballroom is unseen until shot 6, shot 5 turns its head to the mirror wall, shot 7 is lifted and pitched down so the checkerboard reads as a plane, shot 9 comes off the centre line, shot 10 takes the mask wall edge to edge, and shot 11 frames the chandelier's rest and its landing in one shot.
01M2D9P51F4TG635SNBYCQS8ZX848f0f3aa0Browse filesCompare with previousCompare from herefilm: the room reads — velvet stops being white-emissive, the lens stops washing the frame, the eyes light the last shot, and shot 11 stands inside the ballroom
01M1ENTE4PFZNC5APXJP7ZAX71b2801cbd03Browse filesCompare with previousCompare from hereplan: The House Wears a Face — spec, scene skeleton, beat registry, film timeline, five slice stubs --- plan: The House Wears a Face — spec, scene skeleton, beat registry, film timeline, five slice stubs
01M1ECPGG4F2JHQE9JNR66HZWVb4ae075f84Browse filesCompare from here