ROTATE gains the two mirrors: H-FLIP and V-FLIP toggle one at a time and
stay on through the quarter turns and STRAIGHTEN, which makes them compose
with every rotation the strip already offers. ROTATE's own RESET levels the
whole frame, mirrors included.
The flip itself lands last, in screen space, so a mirrored photo is what the
eye sees rather than what the sensor saw; the pixels are copied axis-aligned,
so there is nothing to resample. Session state carries the two flags, so a
reopened photo comes back mirrored.
Also fixes the stamp: a typed PLACE NAME with no GPS fix now prints on its
own (latitude/longitude ride in as NaN), instead of the whole stamp and its
box being skipped for want of coordinates.
The coating's patches were drawn by varying the clump CELL with position:
cell = u * (1 + (grainZone(p * ZONE_FREQ) - 0.5) * ZONE_SWING), the same slow
value noise that picks the patch. A lattice whose cell varies with position
smears instead of resizing: its phase accumulates as
d(phase)/ds = 1/cell - s*cell'/cell^2, and c' is read along the radius from the
picture's own origin, so the second term grows with the distance s from it and
the clumps are drawn out wherever the patch's own cell runs. Measured on one
classic-neg paint (1024px, cell 1.09, the app's own Overlay at alpha 0.5, 64
tiles): with the swing on, the tiles' lag-1 correlation of the raw frame spans
-0.065..0.747 — clumps stretched into smooth blotches beside grain. The design's
+-20% swing cannot do that: the SAME field with the swing forced to 0 spans
-0.057..0.045 across its tiles, and the two-lattice field spans -0.055..0.052
(leica: -0.049..0.523 with the swing on, -0.068..0.024 at swing 0).
The zone now MIXES two FIXED lattices, 0.8x and 1.2x the stock's own cell,
weighted by that same patch noise. A fixed lattice's phase is linear in the
picture, so a patch can only choose how much of each is printed, never how
either is shaped — and the two lattices' beat falls at
1/(1/fine - 1/coarse) = 2.4 cells, 2.6px at the 35mm cell: the pixel scale, not
a line the eye reads. The mix is renormalised by sqrt(w^2 + (1-w)^2), the share
of one field's spread a two-field blend carries, so neither the mean nor the
spread follows the patch: the same classic-neg field reads sd 24.85 against
24.91 and leica 23.74 against 23.74, tile by tile.
The zone still reads what it is for. At preview scale (1600px, cell 1.70, 8x8
tiles of 200px) the mixed field's tiles span rho1 0.082..0.265, ratio 3.233,
against 0.169..0.193, ratio 1.141, with the swing forced to 0 — classic-neg;
leica 0.026..0.188, ratio 7.361, against 0.085..0.105, ratio 1.237. A coarser
patch still prints coarser clumps; it just never prints a stretched one.
Both copies carry it: docker/frontend/shared/utils/grainShader.ts and
src/utils/grainShader.ts (the phone's, which the root web harness imports too).
The field is evaluated once per lattice now, so the grain pass costs 1.94x what
one lattice did — the ratio, not the absolute.
Measured:
_grain-zone2.cjs — the 64-tile lag-1 correlation spread above, three modules
on one paint and one seed: swing 0.4 vs swing 0 vs the mix.
_grain-zone-ck.cjs — the zone's own contribution at preview scale, zone on
against the same module with the swing forced to 0, nothing else differing:
classic-neg tile sd 24.21..24.86 (ratio 1.027), hf 0.769..0.937 (1.219),
rho1 0.082..0.265 (3.233) against sd 29.04..32.95 (1.135), hf 0.834..0.858
(1.028), rho1 0.169..0.193 (1.141); leica rho1 0.026..0.188 (7.361) against
0.085..0.105 (1.237). The swing-off row's higher sd is the renormalisation
of a blend with itself (a and b are one field at swing 0), not a contrast
change in the shipped field.
_grain-fft.cjs — same paint, same seeds, three rolls, 1024px: the mix's top
spectral peak sits at 2.6px (classic-neg, cell 1.09) and 2.8..2.9px (leica,
cell 1.00) against the warped field's 3.0/4.3/6.2px and 2.7/3.8/4.5px — both
within a pixel of the clump cell, neither a coarse lattice.
_grain-bench.cjs — 12.68s per 700px field (one lattice) against 24.60s (two),
1.94x on software CanvasKit.
grain-size-test.cjs 17/0 — the SIZE rule and the readout on the module, both
lattices floored at the target's own pixel, file rho1 0.673, preview rho1
0.003.
grain-stock-test.cjs 53/0 on the deployed build — the stock table, the
halation chain and its ordering, no page errors.
grain-controls-test.cjs 20/0 on the deployed build — the patch claim still
holds there: tile sd 56.58..65.48 (mean 61.9, max/min 1.157), tile mean
spread 0.65, so a coarser patch is still not a brighter one.
_grain-spectrum.cjs (app, deployed, 1600px) — residual autocorrelation peak
0.020..0.021, top peaks at 2.0px@20/110 and 2.5px@51.
tsc: web `--noEmit` clean (the docker build runs it); the phone's scoped
config reports its pre-change baseline, nothing in grainShader.ts.
One honest number: the app-level spectral peak/median rises 4.6..6.1 to 10.2
(classic-neg, sim-classic-neg-g6/g10) because two fixed lattices beat where one
warped lattice spread. It is 20x below the value-noise field this work replaced
(29..35, tiling) and 5x below a lattice (50+), and it sits at 2px, the cell
itself.
ponytail: the field is evaluated once per lattice, so the grain pass costs
1.94x. One evaluation cannot hold two cell sizes; revisit only if a preview
budget asks for the pass back. The 0.8/1.2 rungs (ZONE_SWING/2 either side) are
one working set, not a search.
Verified: `grain-stock-test.cjs` 53/0, `grain-controls-test.cjs` 20/0 and
`_grain-spectrum.cjs` against the deployed build at localhost:8090;
`grain-size-test.cjs` 17/0; `_grain-zone2.cjs`, `_grain-zone-ck.cjs`,
`_grain-fft.cjs`, `_grain-bench.cjs` against the module built from HEAD,
`inversesqrt` still the one call the shader needed to renormalise; web
`tsc --noEmit` clean, phone scoped tsc down to its pre-existing errors.
MONOCHROME GRAIN was one integer knob 0..10 with one meaning, how much. It is now
a strip of three: AMOUNT — the same knob, in half steps — SIZE, a percentage of
the stock's own grain cell (50..200%, so the same number means the same texture
relative to the picture on both platforms), and an inert readout of N/INCH, the
clump count the two knobs and the stock add up to in the print's own terms (300
dpi = 300px of the 1080-wide reference the knob was tuned at).
Emulsion is not one grain size across the frame: the coating settles unevenly.
The field now prints that — the same hash read slowly (ZONE_FREQ = 1/96 cells,
turned off the axes, smoothed so a border between two patches is a slope and not
a seam) swings each patch's own cell by half of ZONE_SWING either way, ±20%.
Nothing in it moves the field's mean: a coarser patch prints bigger clumps, not a
brighter one, which is why the strip can read out one number while the frame
carries a range.
A patch may not swing a cell under the pixel the target can print, or the clumps
are sub-pixel and print as static — aliasing, not a finer emulsion. The shader
takes that floor as a `mincell` uniform beside the cell (u, mincell, seed.xy, in
declaration order): an export passes one output pixel, a preview one device pixel
(1 / PixelRatio), which is the floor the phone's preview already needed.
SIZE is stored as an integer percent so no float noise reaches the recipe JSON,
and it is read by the same two engines that read grain: the web's
grainCell(width, stock, sizePct) and the phone's grainCell(width, minCell,
sizePct). The chip above the strip carries the amount in half steps the way TEMP
carries the kelvin, and the readout moves with SIZE, not with AMOUNT.
Measured:
grain-controls-test.cjs 20/0 — the strip carries grp-grain, grain:amount,
grain:size and grain-inch; the AMOUNT ruler is 0..10 step 0.5, and 3 -> 3.5
moves the frame (sigma 18.53 -> 21.91, new hash) without moving the readout;
10 -> sigma 60.43, 0 -> sigma 0; SIZE 200% -> 130/INCH (sigma 40.06), 50% ->
522/INCH (66.56: under the preview's pixel floor what is printed is static,
not finer grain); the region claim on an 8x8 grid of the flat frame gives
tile sd 51.0..66.1, max/min 1.295, and a tile mean spread of 1.14 — a coarser
patch is not a brighter one.
grain-size-test.cjs 17/0 (was 12/0) — the SIZE rule and the readout on the
module itself: 200% doubles the cell, 50% halves it, the output pixel still
floors the smaller one. 4000px file cell 3.704 against 1.083 device px on a 3x
preview, the old one-dp floor 2.77x coarser, rho1 0.627 against 0.074. The
harness built its own 3-uniform array; it now passes [u, mincell, seed.xy]
like every other caller.
_grain-zone-ck.cjs — the zone's own contribution, at preview scale (cell 1.70,
1600px, 8x8 tiles of 200px): zone on, tile sd 23.22..24.82 (ratio 1.069);
zone off, 24.42..24.79 (ratio 1.015). Nothing else differs.
_grain-ck.cjs — the clump field is otherwise what it was: rho1 0.62
classic-neg / 0.12 velvia, residual autocorr 0.035 against 0.036 with the
swing forced to 0, peak/median 32.3 against 27.6.
_grain-spectrum.cjs (app, 1600px render) — residual autocorr 0.017..0.018,
spectral peak/median 4.6..6.1: the slow lattice adds no peak of its own.
grain-stock-test 53/0, sims-test 31/0, fx-mono-test 15/0, wb-preset-test 33/0,
temp-swatch-test 33/0, wm-font-test 38/0, grain-analog-test 7/0 (its grain
selectors moved to the strip).
tsc: web clean; the phone's scoped config reports exactly the pre-change
baseline (Viewfinder.tsx's own errors, none new).
ponytail: the amount is fractional now, so the two recipe-create forms read grain
through their own half() instead of the int() that would truncate the half the
ruler just spent — every other knob there is still whole. The SIZE knob is one
number for the whole strip: no way to dial a single patch, and no seed control.
The readout is the DESIGN count the field is built on, never a per-patch
measurement.
The field was 20-degree-rotated value noise on a square lattice, three dyadic
octaves at 1 / 0.5 / 0.25 and a sin hash behind it. Value noise prints the
density of the cell's four corners, so every clump sat on a knot of one grid,
and the grid's own repeat — 13 cells, 44px at the 35mm cell — is what the eye
read as diagonal lines. Measured on the old field through the app
(`grain-analog-test.cjs`, `_grain-spectrum.cjs`): off-origin autocorrelation
peak 0.32-0.40, spectral peak/median 29-35, top peaks at 5.4-7.8px.
The field is jittered clumps now, in both copies
(docker/frontend/shared/utils/grainShader.ts and src/utils/grainShader.ts).
A clump lands at a random spot inside its cell — Worley F1 over the 3x3
neighbourhood, `grainClump` — so no two clumps share a grid, and what is
printed is the distance to the nearest: a smooth mound, not one pixel of
static. The hash behind the jitter is sin-free (Hoskins' `p3 = fract(vec3 *
0.1031); p3 += dot(p3, p3.yzx + 33.33)`), because a float sinus whose argument
grows with the picture folds back on itself and is a lattice of its own. The
three octaves are turned to their own angles — 20, 47, 73 degrees — and sit on
0.53 and 0.29 off the dyadic 1 / 0.5 / 0.25, where a coarse octave's cells land
back on the fine one's and stack.
Clumps sit higher and tighter than the value noise they replace — mean 0.569
against 0.500, sigma 0.123 against 0.081, measured — so the sum is put back on
that mean and spread, `n = (n - 0.5685) * 0.52 + 0.5`, before the AMOUNT
knob's own gain. The web keeps its uniforms (cell, seed, per-stock weights,
spread); the phone keeps 0.55/0.30/0.15 and 2.95, which are the web's
classic-chrome row, so the two print the same texture.
Measured on the deployed build: `grain-stock-test.cjs` 53/0 — 35mm still
coarser than 120, the halation chain intact per stock, the "halation follows
the stock" ordering intact, the field still clumped (r1 0.336-0.506) and still
surviving a 2x downscale. `_grain-spectrum.cjs` residual autocorrelation peak
0.02 (was 0.32-0.40), spectral peak/median 7.0-12.7 (was 29-35), top peaks
only at 2-3px periods, the cell scale. `_grain-ck.cjs` at the preview cell
(U=1.481/1.704): rho1 0.093/0.177 against the old 0.505/0.586, acPeak
0.02/0.028 against 0.38/0.467, peak/med 10.1/11.2 against 76/46.5, top peaks
2.0-2.9px against 5.4-7.8px. `grain-size-test.cjs` 12/0.
`grain-analog-test.cjs` 7/0 — with the TEMP pair on AUTO, the field's channel
split measures 0 at a cast of 0, so the grain is exactly monochrome and no
neighbour lag carries structure (max |rho1..8| 0.118).
One honest number: the preview's apparent strength at GRAIN 10 is ~25% higher
than the old field's (sigma 61.3 against 47.9 in that harness), and that is the
PREVIEW, not the field. Step 9 of src/engine/exportEngine.ts sharpens the
preview at alpha 0.5, and the clumps now sit at the cell (~1.7px) instead of on
the old ~5px lattice, so that sharpen bites harder. The field's own composite
spread is 17% UNDER the old one at the same geometry — lab sdRaw 24.9 against
30.0, and geometry-flat where the old one was ~30 everywhere. The 0.52
normalisation was kept rather than re-tuned upward to the app-visible number:
the export path does not carry that sharpen, and a grain tuned to it would
print too strong.
ponytail: the octave angles and the 0.53/0.29 rungs are one working set, not a
search. Re-tune only if a stock's cell is changed again.
Verified: 53/0 + 12/0 + 7/0 grain harnesses, `_grain-spectrum.cjs` and
`_grain-ck.cjs` A/B against the field built from HEAD, `sims-test.cjs` 31/0,
`fx-mono-test.cjs` 15/0, no page errors.
TEMP was a kelvin and nothing else, and the seven presets were seven numbers
the two platforms disagreed about: AUTO 5500, DAYLIGHT 5600, DAYLIGHT -3R
5800, CLOUDY 6500, SHADE 7500 here and 8000 in both recipe-creation forms,
TUNGSTEN 3200, FLUOR 4000. A chip tapped in the recipe panel and the same
chip tapped on the photo printed different frames.
The presets are now the phone's own WB_PAIRS — kelvin AND tint — in all three
places, so every chip lands the same pair on either platform:
AUTO 5500/0 DAYLIGHT 5500/0 DAYLIGHT -3R 5500/-3 CLOUDY 6500/1
SHADE 7500/2 TUNGSTEN 3200/0 FLUOR 4000/3
SHADE read 8000K in RecipeCreatePanel and RecipeCreateModal; it is the WB
tab's 7500K/+2 now, so a recipe created in a form prints what the same chip
prints on the photo.
AUTO and DAYLIGHT stand for one pair, so the pair alone cannot say which of
the two is lit. TEMP is therefore keyed by preset and not by value: `wbChoice`
remembers the chip last tapped (the phone's own wbChoice), and `wbValue()`
answers it only while the engine pair still matches, else the first preset
that pair maps to, else the bare kelvin. `wbLabel()` names that pick on the
chip, which now always carries one: TEMP AUTO at the neutral pair, TEMP SHADE
on a preset, TEMP 6300K on the app's own default recipe where a hand-dragged
ruler landed. Measured on the deployed build (`wb-preset-test.cjs`, 33/0):
AUTO and DAYLIGHT print the same frame to the level, DAYLIGHT -3R moves the
green away from DAYLIGHT at the same 5500K, the ruler warms monotonically
across TUNGSTEN 0.525 / FLUOR 0.730 / AUTO 1.000 / CLOUDY 1.141 / SHADE 1.295,
a hand-drag to 10000K names the chip `10000K` and warms the picture R x1.183
B x0.793, SHADE tapped after that drag restores the pair 7500/2 and prints the
same frame the first tap did (R 156.8 vs 156.8), and a temperature drag leaves
a preset's tint standing (FLUOR's +3).
kelvinToRGB itself was the other half. It took the ratio of the sRGB-ENCODED
blackbody colours and square-rooted it, which measured R x1.06 / B x0.89 from
5500K to 10000K — a shift a swatch shows and a sunset does not. White balance
is a gain on LIGHT, so the ratio is taken in linear light now
(`planckianLinear`) and tamed by a new `KELVIN_TAME = 0.5`: the same 10000K
moves the frame R x1.16 / B x0.76, and 2500K its mirror, which is what a
camera does with its WB set 4500K off the scene. One constant scales the whole
ruler and both platforms carry the same one. Measured through the app
(`_temp-probe.cjs`, gains against 5500K): 2500 R x0.569 B x1.640, 4000 R x0.866
B x1.197, 6500 R x1.057 B x0.940, 10000 R x1.140 B x0.759 — the readout is
compressed against the raw gain because the cast lands on encoded, clipping
pixels, which is the reason for the tame in the first place.
ponytail: no scene meter, so AUTO stays the neutral 5500K/0 pair and is a name
for it, not a measurement. Add one when the engine reads the frame.
ponytail: KELVIN_TAME scales the whole ruler both ways. Split it into a warm
and a cool constant only if the two ends are ever asked to move apart.
Verified: `wb-preset-test.cjs` 33/0 and `_temp-probe.cjs`, `sims-test.cjs`
31/0, `fx-mono-test.cjs` 15/0, `wm-font-test.cjs` 38/0 against the deployed
build; web `tsc --noEmit` clean, the phone's scoped check down to its two
pre-existing `skiaImage.ts` nulls.
The camera preview, the library preview and the exported file each carried
their own copy of the grain shader, and each sized it off the rect it happened
to be drawing on. Two things were wrong with that. The field was a single value
hash, so it looked like digital static rather than film: measured on the noise
plane, neighbour correlation was flat and the lattice sat on the picture's own
axes; and it was the same field in every session, so two photos from two
launches printed identical grain. The size rule was wrong the other way round:
canvas units in Skia are dp, so on a 3x phone the floor of one dp pinned a
390dp-wide preview to a THREE device-pixel cell while the file printed a
1080th of its width — 2.77x coarser than the picture it was previewing, which
is the preview-vs-file mismatch all over again, just the other direction.
Grain now lives in one module, src/utils/grainShader.ts, imported by the
preview (both surfaces), the phone's export and the web engine, so the three
cannot drift. The noise is three octaves of value noise over a density that is
the mean of three hashes — a bell, the way an emulsion's density swings —
turned 20 degrees off the axes, seeded once per app launch, so a session prints
one roll and the next launch prints another. Only coarser octaves: a finer one
lands under the pixel and the clumping is lost outright. The cell is one
1/1080th of the PICTURE's width, floored at one pixel OF THE TARGET — an
export's own output pixel, a preview's device pixel (1 / PixelRatio.get())
instead of a whole dp.
Measured on the plane itself with the real module (grainSize harness): the file
at 4000px and a 390dp camera preview at 3x agree on the same fractional lag,
rho 0.017 against 0.011, while the floor this replaces reads 0.329 — visibly
coarser, as claimed; both keep the spread the AMOUNT knob was tuned on
(sigma 57.8 / 57.9) and still clump (rho(1) 0.91 / 0.39); R, G and B never move
apart, split 0. GRAIN_SKSL compiles on the Skia runtime, and the compile is now
swallowed into null at module scope like the other three effects, so a shader
this app cannot compile costs the grain rather than a launch.
Not ported: the Kotlin grainOverlay probe (per-pixel, path is off by default).
A second black-and-white stock, so it gets its own baseFilter rather than
borrowing Acros': the PHOTO STYLE chips are keyed by baseFilter and the two
greys have to sit side by side.
The look is the B&W MIX plus a push at both ends. The mix rides the matrix — a
non-BT.709 row set (0.38/0.56/0.06, identical rows, sum 1.00) so a red roof
reads bright, a blue sky deep, and the separation is contrast before any curve.
The push rides FILM_TONE (shadow -0.32, highlight +0.26) so the ends move
without touching the midtones, and the midtone slope is SIM_CONTRAST_BIAS (4
contrast units). Every knob still opens at neutral: a sim is colour and tone
only.
Both B&W stocks being mono is now asked once, through isMonochromeBase, so the
colour-only stages (saturation, white balance, the R/B fine-tune) and the
chrome/hue-mixer gate can never half-apply to one of them.
Mirrors the web app's rail (c403dd0).
Bound the geotag fix with a 2 s deadline and prewarm it at launch, queue a
shot pressed before the camera session binds and replay it, pass the options
argument capturePhotoToFile requires, save through the native path so the
media permission is not needed, and release the Skia surface and bitmap after
every export. Store the JS bundle uncompressed and load the export pipeline on
demand. Note in RELEASE_NOTES.md that none of this moved the tap threshold.
0.5x is a bare Camera2 session, so CameraX never saw the flash setting and the ultra-wide still came off unlit. The native module now reads FLASH_INFO_AVAILABLE off the lens, maps off/auto/on to CONTROL_AE_MODE plus FLASH_MODE_SINGLE, and starts an AE precapture so the HAL can meter the flash before the frame. The viewfinder hands its flashMode down beside the shutter, with a Zap control that only appears where a unit exists.
Library CROP now resizes the band in photo fractions, the way FREE always
did, instead of panning a re-scaled photo under a screen-sized rectangle:
what the band encloses is the export, with no zoom transform left to undo
and no aspect that only holds on a square photo. The band is laid out
below the top strip, whose height TopBar now reports, so it can never
peek under the header. The grain hashes a cell of width/1080 instead of
the raw pixel grid, so a 4000px export is no longer ~4x finer than the
preview that tuned it.
The ultra-wide preview is a bare Camera2 surface: it never goes through
Skia, so the software EV gain that carries exposure on the 1x preview had
nothing to act on and quick EV / the LIGHT-EV slider did nothing at 0.5x.
The lens itself now takes the bias, as CONTROL_AE_EXPOSURE_COMPENSATION on
every preview request (session builder, metering rebuilds, and a setEv call
the viewfinder pushes whenever the value changes).
The still is left alone: capture() pins the compensation back to 0 and
waits on an AE precapture trigger, so the file stays unbiased and the export
applies its software gain exactly once, as it does for the 1x path. The
hardware range clamps the bias where the lens offers less than the app's
+/-3 EV.
- One free Play app (com.locphamtran.recipescamera) with an expo-iap PRO
unlock, replacing the paid apk for the store channel
- Android SEND/SEND_MULTIPLE share target feeds photos into the editor,
on cold start and on resume
- BACK button in the TopBar and press-and-hold PEEK replace the chip row
- NOISE REDUCTION and SHARPENING accept -10: a negative SHARPENING softens
the frame before the grain pass, a negative NOISE REDUCTION re-grains it,
in the Skia engine, the live preview and the Kotlin export path alike
Rename LIGHT GLOW to HDF EFFECT and gate it behind PRO: LITE renders the
chip greyed out with a PRO tag and cannot open its slider; restoration of
older settings that carry hdf > 0 still blocks export in LITE.
Play/licence pass:
- drop unused permissions (ACTIVITY_RECOGNITION, READ_MEDIA_AUDIO/VIDEO,
SYSTEM_ALERT_WINDOW) and requestLegacyExternalStorage
- rename the Leica sims to LC STREETLIFE CLASSIC/VIVID, drop the brand from
comments; fix app name text, canvaskit copyright, LICENSE owner
- remove expo-image-picker and expo-sensors (no JS import), delete dead
exports (computeMatrixTint, cinemaSeasonName, base64 helpers,
formatCoordinate)
- versionCode 3 / versionName 1.2.1, drop the personal email from About
- add docs/privacy-policy.html for the Play listing
Size: R8 + shrinkResources + bundle compression + ABI trim, 161 MB -> 76 MB.
Freeing the FAVORITED tab for LITE put no ceiling on the list, and a list
of your own picks is what PRO is selling. One star is free, the second
names the build; un-starring stays open so the list can always come back
under the limit.
LITE used to hand back the RECIPESCAM mark instead of the user's own
watermark and treated every saved recipe as PRO. That is the wrong half of
the product: the looks the user dials in are the reason to shoot, the
printing and the paid extras are the reason to pay.
What LITE now owns outright: the eight film sims, the bundled recipes, up
to three recipes of its own, its own watermark text (in one basic face,
the platform families stay PRO), the plain frames, and FAVORITED. What
needs PRO: the three printed frames, the GPS stamp, the other fonts, and
the next recipe past the limit.
- entitlement: `Look` is {frame, gpsWm}; `isFreeRecipe` is gone, the
recipe gate is a count now, decided in App (`liteSaveBlocked`).
- export: a LITE export draws the custom mark AND the RECIPESCAM mark, so
`liteMark` joins `watermark` in the options and the stamp block loops
over both with one set of bounds.
- panel: past the limit the strip's trailing chip is a `+` carrying the
PRO tag, which names the build instead of opening the form.
LITE lets you pick any preset, frame or watermark and preview it; only the
export is held back, and nothing on screen said which chips those were until
you hit the alert. Each such chip now carries a small amber PRO tag at its
top-right corner.
ChipDef gains an optional `pro` flag, set on the three printed frames, on
recipe chips (bundled or user-made, via isFreeRecipe), and on the custom
watermark chip. The tag renders only where `!IS_PRO`, so it is invisible in
the PRO apk.
The set is two apks and the unlock key was a second, weaker lock on the
same door. A verifier that runs offline has to carry its secret inside
the lite apk, so anyone who unpacked it could mint a key. Nothing in the
lite build can flip the gate now.
- entitlement.ts keeps the gate (which look is PRO, what an export
carries in lite) and exports IS_PRO straight from the build flag.
- SETTINGS loses the key field; the plan chip still says LITE / PRO
ACTIVE and the helper text says what lite cannot export.
- The upsell alerts just name the build that has the feature.
- tools/keygen.mjs goes with the verifier it fed.
Both flavors build from one JS tree: `lite` hands out the free apk
(com.locphamtran.recipescamera), `pro` sells as-is (same id + .pro, label
"RecipesCam Pro"). Release is signed from android/keystore.properties, both
files gitignored - lose either and no later build can install over this one.
The tier reaches JS through src/provariant.ts, which tools/set-variant.mjs
rewrites before each build (`npm run apk:lite` / `apk:pro`); the gradle flavor
alone would ship a Pro apk that thinks it is Lite.
The dev scaffolding goes: the three src/dev probes and their App effects, the
EXPO_PUBLIC_CAPTURE_MODE [CAPTURE] timing log, and the -PdevtestSuffix
side-by-side install hook. 'balanced' stays the shipped capture mode, which is
what those probes measured on the 12S Ultra.
Looking at the picture: WB now converts kelvin through the Planckian locus into
a luma-normalised gain (unity at 5500K, symmetric tint) instead of the old
one-sided push; LEITZ STREETLIFE splits into CLASSIC and VIVID; selecting a
recipe loads every knob it carries over the defaults and stays editable, and
RESET hands the panel back to the sim's own values. The in-app library picker,
the ultra-wide native module and recipe share/import land here too.
CLARITY and LIGHT GLOW existed only on the camera worklet and the export
engine, so a photo opened from the library ignored both. They now run in all
three library preview branches (libPhoto), and the preview clips them to the
drawn image - the export engine always clipped its bloom to the canvas, the
preview drew it over the whole rect, so the bloom bled past the photo onto the
black letterbox and over the frame's mat.
Dragging the date bar fired a window query every 120ms while one query takes
over a second on a 12S Ultra, so ~20 queries queued behind each other, the
grid froze for the length of the drag and the stale one landed last - the
grid settled on a date the finger had long left.
The window now opens on both sides: a page above the target, so a list parked
mid-library still has content above it and a downward drag does not die on
offset 0, plus the target page and everything older. jumpTo coalesces the
drag into one query and parks the grid on the row the target date sits on;
scrolling back up to the top pages newer photos in and corrects the offset by
hand, because maintainVisibleContentPosition keeps the offset glued to index 0
and walked the grid forward in time as it paged.
LITE previews every look and feature but export/render is blocked while a PRO
look is in use, and every LITE export carries the RECIPESCAM mark. Keys are
generated offline (FNV-1a checksum over a fixed secret, 16 chars from a
no-ambiguity alphabet) and activate one machine or a thousand: nothing is bound
to a device, so revoking a leaked key needs Play Billing plus server validation.
CREATE, FAVORITED, the star chip and the TopBar + SAVE now answer with a PRO
prompt instead of a silent no-op; CREATE still drafts and applies the look so
LITE can try it, it just cannot persist.
The SETTINGS card also lifts above the keyboard now: RN Modal is its own dialog
window, so the IME never resizes or pans it and the UNLOCK row was sitting under
the keys.
Pressing and holding on the photo with a LIGHT/WB/FX panel open asks App
for a preview drawn without the edit that gesture just made; lifting the
finger asks for the edited look back. Pegged to one gesture, not one
pointer-move: App only re-snapshots the 'before' state after a 700ms
pause, so the whole drag compares against its own starting point.
Preview-only - the sliders, RECIPE and the export keep reading the real
adjustments. Moving past TAP_SLOP, a second finger, or a release all end
the peek, and a peek is not a tap, so tap-to-focus and double-tap zoom
are untouched.
- FRAME tab now pinches the whole view (frame + photo + watermark) in both
camera and library; one finger pans that view. Preview only: the exported
file is identical at 1x and while the view is magnified.
- Watermark pinch/drag stay relative and keep priority while the WATERMARK
row is open, so they still work after returning from the FRAME tab.
- Exports go through a native MediaStore insert into DCIM/Camera, falling
back to expo-media-library when the native save fails.
The watermark chip used to be the only GPS control, so turning the watermark
off also stopped the photo from carrying coordinates. Split the two: a GPS
master switch in Settings decides whether the location is read at all, and the
chip decides whether it is drawn on the photo.
0x889d is also written as UTF-8 now — a place name with accents ("TAM KỲ, ĐÀ
NẴNG") was going out latin-1 and reading back as mojibake.
The framing tool matches Make/Model against its known-device list and prints
the place name; a CameraX capture only carried the raw model codename
(2203121C) and no 0x9a00/0x889d, which is what a stock camera photo always
has. State the market name plus the geotag locality, and leave a source that
already carries 0x9a00 untouched so a library photo of another device is
never relabelled.
HyperOS Gallery reads EXIF 0x889e (a JSON blob) to build its watermark
frame; without it a photo fails with "cannot recognize the parameters".
CameraX/HAL never supplies that vendor tag, so a capture from this app
lacked it while a stock-camera photo carried it.
Read back the same props the stock camera derives the blob from
(ro.product.device / ro.product.marketname / Build.MANUFACTURER) via the
native module, synthesize the blob in exifWrite, and write it only when
the source has none and the device is Xiaomi/Redmi/POCO - a source blob
is never overwritten.
Match the layout the sample camera file uses: a fourth Interoperability IFD
(0xa005 = R98/0100) and the baseline tags the emulator HAL omits (ExifVersion
0230, FlashPixVersion 0100, ComponentsConfiguration, ColorSpace), plus the
OffsetTime/OffsetTimeOriginal/OffsetTimeDigitized tags for shots this app
timestamps. Source-owned tags are still kept as-is.
Also fix swapToLittleEndian: Buffer.slice() aliases, so the big-endian swap was
mutating the caller's bytes (a Xiaomi source read back as ISO 36865 instead of
400). Use new Uint8Array(data).
- FRAME/CROP: choosing a ratio shows the amber band (border + 0.55 dim);
APPLY collapses the preview to the crop rect with an opaque mask and
removes the amber stroke; RESET returns to 'none'.
- Viewfinder: libCropView (k = min(vw/dw, vh/dh)) + cropScreenPx drive the
mask/band, libViewMatrix folds the crop transform into the image groups,
libCropTouchStyle keeps touch mapping aligned.
- Persist cropApplied in the session snapshot and restore it only when it
still matches cropRatio.
- Add PLAN-2026-09-09.md with the measurements.
tsc --noEmit unchanged at 14 pre-existing errors.
EXIF Orientation only knows 0/90/180/270, so a library still carries no
record of how the camera was held. The sensor had nothing to offer.
AUTO now measures the dominant line in the picture itself:
src/utils/horizon.ts runs a shear-projection search (coarse 1 deg over
-45..45, then a 0.5 deg refine) on a <=256px thumbnail, bails when no
line's score beats 3x the median, and returns the tilt in degrees.
App applies photoStraighten = -tilt, so preview and export share one
number exactly as the slider did.
Removes expo-sensors wiring, the horizonRoll state, effectiveStraighten
and the bubble-level overlay (autoRoll prop) from App/AdjustmentPanel/
Viewfinder. expo-sensors stays in package.json.
AUTO turns the amber level line on over the photo, but the line only left
when AUTO itself was switched off: switching to another frame chip, another
parameter or another rail tab kept it painted on the picture.
Gate the line on the ROTATE context instead: AdjustmentPanel reports whether
the ROTATE strip (or its STRAIGHTEN row) is open, App keeps autoRoll non-null
only then, and the frame chips now close the strip like every other chip
does. Verified on emulator-5554: AUTO on with the strip up shows the line
(row 1199-1203, 594 px); tapping LIGHT and coming back leaves it off.
Four follow-ups on the strip added for the quarter turns / straighten / AUTO.
AUTO is "snap to the sensor horizon": it now drops the manual STRAIGHTEN offset when it is switched on, so enabling it after a hand-set angle really levels the photo instead of leaving the angle untouched (the chip goes from "ROTATE 0 · +22°" to "ROTATE · AUTO").
The "<" back button on a slider row returns to the strip the row was opened from (STRAIGHTEN to ROTATE, COLOR TEMP to TEMP) instead of closing everything, and picking another parameter — a frame chip or the global RESET — closes the open row, so the straighten slider no longer outlives the parameter it belongs to.
Opening another photo from the library resets the turn, the fine straighten angle and AUTO with it: a new photo starts level.
The ROTATE strip gains an AUTO chip that reads the device tilt from expo-sensors and feeds it into the straighten angle while the library is open, so the export is levelled to the horizon. It sits right after RESET, combines with the manual STRAIGHTEN offset and with the quarter turns, and both RESET paths clear it.
The accelerator is sampled every 100 ms only while AUTO is on and the library is visible; near-flat and past 45 degrees readings are ignored so the level does not chase noise. AUTO rotates the image by +roll (a +10 degree emulator tilt exports 10 degrees clockwise), which is the sign that matches the preview.
A level line overlay is drawn in the preview: amber within one degree of level, white otherwise. Adds the expo-sensors dependency, so a native rebuild is required.
Two fingers scale the live feed inside a polaroid card / wall opening and
one finger drags it, both clamped to the window, and the shutter passes the
result to the export as frameWindowZoom so the printed crop matches what the
viewfinder showed. The crop centre is mapped through the same out-space ->
raw transform as the window rect; feeding it straight to the raw plane
swapped the axes on a rotated sensor.
RESET was look-relative: it only put the sliders back on the values the
active recipe ships with, so the look itself (a saved recipe, a film sim)
survived the tap. It now restores the state a clean start leaves — the
PROVIA startup look, every parameter neutral, frame off, the custom and
GPS watermarks back to their defaults — and it sits on every tab, pinned
outside the scrolling chip row so it can never scroll out of reach.
Mode and aspect ratio stay put: they frame the shot being worked on, not
the look.
3dtours
admin