The panel shared one column between the two marks, so GPS's colour, its two
switches and its hand-typed place stood open beside the custom mark's text,
colour and size whether or not either mark was on. The two are now collapses,
one per mark: the header chip is the section, and that mark's own controls sit
under it. What opens a section is the mark itself — GPS WATERMARK ON opens
GPS's controls, CUSTOM WATERMARK ON opens the custom mark's — so there is no
new state and no way for a panel to disagree with the pixels.
Both marks gain the FONT strip the phone has had (TEXT FONT for the custom
mark, FONT for GPS, whose stamp the phone also lets you set a face on). A
browser has no font service, so the list is exactly what the bundle carries:
the site's two self-hosted families, Inter and Fraunces (SIL OFL), their latin,
latin-ext and vietnamese woff2 subsets decompressed, pinned to weight 400 @
opsz 14 and merged into ONE TTF per family — drawText has no glyph fallback, so
a family mapped to only the latin subset would print a Vietnamese place name as
tofu. DEFAULT stays the bundled Cousine face, which is what every existing
session and every mark without a family prints.
Two engine bugs came out of it. CanvasKit 0.42's Font.getGlyphWidths passes its
output pointer where the wasm export wants the bounds pointer, so every glyph in
a run comes back holding one identical, rounded width — at 64px on the merged
Inter face, 'H' and 'i' both answered 42, while hmtx says 0.743em and 0.242em,
and a box measured off it was 27% too wide ("Hà Nội 09/23" 510px against a true
403px). The shim now rebinds it with the pointers in the order
_getGlyphWidthBounds reads them, and the stage's boxes measure with linear
metrics, which land on hmtx exactly (403.28px against 403.28; hinted is 407).
And CanvasKit's TypefaceFontProvider.matchFamilyStyle answers null for every
style shape this binding accepts, so a name registered with it never resolved —
the shim keeps its own registry keyed by family name instead.
Measured: tsc clean; the engine harness on the merged faces 26/26, including the
registry's advances against hmtx (Inter 6.3013em, Fraunces 6.3475em); the
deployed app under Playwright 38/38 over the two collapses and both FONT strips
— each mark's controls appear only with its own mark on, the DEFAULT/INTER/
FRAUNCES box widths match hmtx, the baked ink fills the box, the top edge
re-hangs off the new ascent (Inter 0.96875em against Cousine's 0.8325em, 3.4px
at this size) with the left edge fixed, and UNDO round-trips. Opening a section
narrows the stage by 168px with no window resize (955px -> 787px), so the stage
now re-measures its drop boxes off a ResizeObserver on the frame and the
picture rather than on the next render.
Not ported: the phone's GPS watermark still prints in the bundled face only
(no emulator here to verify a phone-side font strip), and the FONT options are
not behind the PRO gate the way the phone gates non-default families.
GRAIN was one value hash per cell, sampled straight off the pixel grid: a
square lattice at the picture's own axes, the same field in every session and
in every photo, and — measured on a flat gray frame, where every deviation IS
grain — a spread that was flat rather than emulsion-like (neighbour
correlation rho(1) = 0.183, so half the noise was one pixel wide).
The noise is now an emulsion. Three uncorrelated hashes are averaged into a
density (a bell, the way an emulsion's density swings, instead of the flat
spread of a single hash), that density is read as value noise with a
smoothstep cell, and the result is three octaves of it — the cell, then 2x and
4x that cell — over a lattice turned 20 degrees off the picture's axes, so no
grid shows through. Only coarser octaves: a finer one (tried 2.043x, 0.72px)
goes sub-pixel and rho(1) falls to 0, i.e. back to static. The whole domain is
offset by a seed rolled once per page load, so two visitors never print the
same clumps while the preview, the compare copy and the file of one session
still print the same roll.
Same flat 3000x2000 frame, preview render 1600x1067, GRAIN 10: sigma 52.53 ->
50.15 (the 2.95 gain keeps the spread the AMOUNT knob was tuned against, since
the repo notes the slider was calibrated on the old field), rho(1) 0.183 ->
0.275, and the three channels still move together — the 6.9 of sigma 50 that
is left over is the Overlay blend meeting the frame's own tint, the plane
itself is one gray value in all three. Re-rendering the same frame draws the
same clumps byte for byte; a reload rolls a new seed and a new field at the
same strength.
A photo restored from the session reaches the stage while /me is still in
flight, so the account reading that put it there saw `pro` as false, the place
lookup was skipped, and reloading a photo left the stamp with bare coordinates.
Naming now happens in one effect that watches the position itself: the first
moment it is on the stage unnamed and the account is PRO, it gets a name,
wherever it came from. A position typed in by hand earns its name too, and a
guest's photo is named the moment they sign in.
The two call sites that used to ask for the name — in locateMe and adoptPhoto —
are gone, since the effect covers both.
The GPS mark printed Date.now(), so a photo taken in 2019 carried the day it
was opened. It now prints the frame's EXIF date — DateTimeOriginal, falling
back on CreateDate then ModifyDate — wherever the position came from:
- readCapturedAt() reads the date off the file, and readGps() uses it for a
position found in the same EXIF.
- adoptPhoto holds it in its own state, so a frame with a date but no position
still stamps the date when the position is typed in by hand.
- The device's own position stamps it too. That path runs inside adoptPhoto,
where the render still holds the previous photo's date, so locateMe takes the
date as an argument rather than reading state — the panel's own button, which
has no such date to hand, passes none and reads the state as before.
A file with no date at all still falls back on the visitor's clock: there is
nothing else to believe.
The WebGPU execution provider has no PReLU kernel. The model is 34 convolutions
with a PReLU after every one of them, so an export that took the GPU path was
split 33 times: each activation came off the chip to be activated on the
processor and went straight back, a 64-channel map in both directions, per tile.
A machine with a good graphics chip was not exporting any faster for having it.
PReLU(x) is exactly Relu(x) - slope * Relu(-x), and Relu, Neg, Mul and Sub the
provider does implement, so scripts/realesr-gpu.py writes the 33 activations out
as those four and drops the slopes nobody reads any more. The model file is the
output of that script, not the file as published.
One 256x256 tile through the model before and after, on a WebGPU session: the
runtime no longer reports nodes left off the preferred provider (it did, once,
before) and the processor path answers bit for bit what it answered before. The
warning itself cannot be switched off from here - env.logLevel is read when the
runtime module initialises, before any of this runs - so the graph was fixed
rather than the lines hidden.
onnxruntime hands a webgpu session whatever adapter the browser picks by
default, which on a laptop with both is the one built into the processor: the
export then waits on the slow half of the machine for no reason. The runtime
reads `env.webgpu.powerPreference` when it builds the webgpu session, so set it
to high-performance; a browser with nothing to honour the preference with
still falls back to the threaded wasm path exactly as before.
The model's own factor is 4 and the export's target is some number of pixels,
and the two were never reconciled: a 2400x1800 photo exporting at 4K was run
through the model at 4x — 9600x7200 of invented detail — and then three
quarters of it were thrown away by the draw that lands the file on 3840. The
arithmetic was the whole wait. Measured on the wasm path, one export: 153.2s.
The photo is now resampled once to `targetLongest / 4` before the model reads
it, so the model still answers at its own 4x and the answer is the size the
export asked for. Same 2400x1800 to 4K: 42.7s, 80 tiles of model for 20. Half
the photo's pixels is the floor — below that the model is no longer enlarging
the picture, it is drawing a new one from memory — and the ceiling is the
photo's own size, so a gain past 4 behaves exactly as it did.
Nothing in the finished file gives the smaller input away: the 6px stripes come
back at full contrast (254.9 vs 254.8), the black-to-white step lands on the
same pixel (x=625 in both) and rises in 1px instead of 3.
The 32MB of runtime and model are also fetched, and one 16x16 tile pushed
through the graph, when the export menu opens rather than after a size is
picked: the visitor waits for the pixels, not for the download.
crop 1:1 2400x1800 to 4K: 155.8s -> 52.6s, crop 3:4: 153.0s -> 54.1s.
The row below the photo carried CLEAR, OPEN and SAVE ORIGINAL but never said how
big the picture was, so the only way to learn the resolution was to open the
export menu and read the hint there. The size now leads that row, before CLEAR:
the file's own pixels turned by the quarter turn and cut by an applied crop, so
it is the number an export at the photo's own size writes. A live crop does not
move it (nothing is cut yet), STRAIGHTEN never does (the rotated rectangle is
fitted back inside the same pixels), and the guest tier's 2048 cap is still only
announced in the export menu where the file itself is capped.
Verified end to end in photo-dims-probe.cjs: 2400x1800 opens as "2400 × 1800",
a 90 deg turn reads 1800 × 2400, an applied 1:1 crop reads 1800 × 1800, and the
export writes exactly that file.
The polaroid and the wall frame both draw the photo into a window that is not
the size of the photo — the instant print shrinks it to 0.898x, the wall frame
cover-scales it by max(winW/pw, winH/ph), which is 1.47x up for a 2400x1800
source. A plain drawImageRect is nearest on CanvasKit, so that resize dropped
the edge back onto the output pixel grid: a 5 deg straighten inside a frame
exported 59.8% of its rows with the crossing pinned to the same pixel as the
row above (61.1% in the wall frame), while the same photo without a frame came
out at 55.8%.
Both draws now go through drawImageRectOptions with FilterMode.Linear, the same
call shape the straighten draw already uses. Measured on a 2400x1800 hard-edge
fixture at 5 deg, exported at the photo's own size: polaroid 59.8% -> 37.5%
flat rows, fracStd 0.336 -> 0.184; wall frame 61.1% -> 25.7%, fracStd 0.467 ->
0.210 — and the residual matches the 0.202 of the unframed export, so the
window costs nothing beyond the resample underneath it. A 45 deg edge printed
through the instant frame at 0 deg is unchanged at 0.0% flat rows.
The FRAME tab's fine rotation drew the photo through canvas.rotate() +
canvas.scale() with a plain drawImage, which CanvasKit samples with
nearest: the edge landed on the same pixel in every row, so a rotated
edge came out as 1px steps every 1/tan(angle) rows. Measured on the 30deg
export of a hard black/white edge: 42.3% of rows repeated the previous
row's edge position, the step across the edge was 252.9 of 255, and there
were no intermediate pixels at all.
Only the *Options/*Cubic call shapes take a sampling option, and
drawImageRectOptions exists in RN Skia too, so the shared renderer can use
it unchanged. The same export now moves the edge in every row (0.2% of
rows repeat, 0.35 intermediate pixels per row) and its edge step drops to
222. Cost: the filtered draw takes 0.35s against 0.24s for the 1600px
preview copy and 2.0s against 1.4s for a 12MP photo, once per render.
Preview and export share the function, so both change together.
Saving into the account's own folder has always been the account's act —
the button opened the way in and the API answers an unproven address with
a 403 — but nothing on the button said so, so it read as a button that
quietly did nothing. It now wears the same PRO marker the chips do, and
only while the folder is not the visitor's.
MY PHOTOS is that folder's listing, so the tab is only offered once an
account can hold one. A guest loses the tab entirely rather than opening
it on an empty folder that could never fill; an account that has signed
up but not proven its address keeps the tab, and the tab keeps offering
the way to prove it.
The last three PHOTO STYLE looks (B&W HIGH CONTRAST, LC STREETLIFE
CLASSIC and LC STREETLIFE VIVID) and the whole mixer now belong to the
account, the way PRO frames and the geotag already do: the chip wears
the PRO badge, a guest who picks it is shown the way in, and the look
stays off. The HSL tab keeps its place in the rail but offers the one
PRO chip while locked, so the tab itself is not a dead end; a look that
arrives without the chips — an imported .recipe, or a photo saved
before the gate — is still caught where the gate bites, at export.
STRAIGHTEN's scale turns with the wheel, one degree a notch, because
the ruler is where the angle is being judged and reaching for a slider
elsewhere loses the thread. The listener is native and stops the notch
before the stage sees it, so the photo does not zoom under the pointer.
The scale gives up its opaque card, its blur and its shadow: the frame
it is levelling has to stay readable through it, so legibility comes
from a text shadow on the heading and a drop shadow on the graduations
instead.
The export menu measured the photo off the 1600px preview copy, so a 2400px
photo was believed to be 1600px across: the hint named the wrong size, the
model was asked to upscale a photo that already had more pixels than the
target, and a guest's 2048 ceiling was skipped because 1600 never crossed it.
A committed crop made it worse — the crop's longest edge was taken from the
wider side of the crop rect rather than the side the frame actually keeps, so
a 2400x1800 photo with the default 0.8 frame was called 1280px and ran the
model over 80 tiles (158.7s) to reach 2K.
The photo's own dimensions are now read off the original bytes, and the crop's
long edge is the same axis-aware fraction the stage already uses. The export
asks the model only when the photo itself is short of the requested size, or
when the crop would have to be stretched past 1.5x to get there; otherwise it
resamples — down, or a hair up to make up for the crop — which is what a photo
that already holds the pixels deserves.
Measured, wasm path, 2400x1800: no crop at 2K went 2.7s/2400px (wrong size) to
3.4s/2048px, the default 0.8 crop went 158.7s/80 tiles to 3.5s/no model, and a
1:1 crop went 2.5s/1800px to 4.0s/2048px. A 1200x900 photo cropped to 1:1 and
exported at 2K still runs the model (2048 from a 900px crop, 40.3s), and the
superres suite is unchanged: 640x480 to 2K/4K/custom still comes out exact,
with the model's 16.6 edge energy against bilinear's 4.8.
The super-resolution export ran single-threaded because the site was not
cross-origin isolated and the runtime had no SharedArrayBuffer to spread a tile
over. nginx now sends COOP and COEP — on the document, and on the script
responses a nested worker fetches, which Chromium checks the same way and blocks
as `coep-frame-resource-needs-coep-header` without them — and the loader asks
for `min(8, hardwareConcurrency)` threads whenever the page is isolated, falling
back to one if the headers ever go missing. A worker script is also why the
landing's QR image needed `crossOrigin`: COEP refuses a cross-origin image that
did not opt in with CORS.
The unpack was the other half. Each tile was clamped a channel at a time and
painted whole, padded ring and all; it now writes straight into the
Uint8ClampedArray, which clamps and rounds on assignment, and skips the ring
rather than drawing it and clipping it away.
640x480 to 4096: 39.0s to 16.2s. 1000x750 to 4096: 89.6s to 30.9s. One 256px
tile through the model: 6.8s to 1.8s. Measured on the wasm path — the test
browser has no GPU adapter — so a WebGPU export, still per-tile inference, keeps
its own times.
The chip lands after ACRIPES and is a mono stock of its own, so it gets its own
baseFilter ('mono-high-contrast') rather than borrowing Acros': the PHOTO STYLE
chips are keyed by baseFilter, and the two greys must 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) next to the sim's own exposure bias. The knobs stay 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, R/B fine-tune, chrome, hue
mixer) and the MONO strip label can never half-apply to one of them.
The card carried studio copy — SUNSET GLOW, a made-up warmth/grain/bloom line
and a creator handle borrowed from a testimonial — no matter which still the
arrows landed on. Stepping the frame changed the picture and the code but left
the text behind, so the card described a photo it was not showing.
The card now reads the frame's own labels: the tagline over the still and the
title and ISO/grain line under it are the ones the uploader saved with the
photo, the same three the film strip prints. The fabricated creator/imports
line goes with them, since nothing behind it was real.
The marquee loops by translating the track by half its width, so the first half
has to be at least as wide as the window. A short reel — seven stills, about
1780px — covered a 1440px window but not a 1920px or 2560px one: the strip ran
out of frames before the loop restarted, and the band on the right stayed blank
until the next pass drifted in.
The track now measures one frame's pitch and repeats the reel until a half
covers the window, re-measuring on resize. One still in the strip is repeated
enough times to loop cleanly on its own; a reel already wide enough is left at
a single copy, so nothing is duplicated without cause.
The landing's tester preview has had a ‹ › pair for a while; the custom recipe
creator and the QR recipe card still showed one arbitrary frame from whatever
the curator tagged for them. Both now cycle their own slot the same way, from a
random start so the page does not look identical on every load, wrapping at
both ends.
The QR card's payload is not decoration: it is the link of the frame on screen,
so the code under it is redrawn from the same frame the arrows land on. A slot
holding fewer than two stills gets no arrows, since there is nowhere to step.
The tester's arrows move to the shared FrameArrows component, which is what the
two new pairs use — one implementation, three slots.
COLOR RECIPES, DOWNLOADS and STORE RATING were flat grey under their numbers.
They now sweep from the theme's own accent to the film red and are cut out of
that gradient, the same treatment the headline above them gets. Because the
sweep starts on var(--accent), the colour group in the Theme menu still
retints it; the light theme starts from the darker ink accent so the labels
keep their contrast on white.
Six bundled sample negatives stood behind eight built-in looks, and they were
the only frames on the page nobody had uploaded. They are gone, along with the
REEL table and the SAMPLE() helper that pointed at them.
The film strip is now exactly the photos the curator put in the strip slot,
repeated once for the marquee loop; a section whose slot holds nothing simply
draws no frame instead of falling back to a stock photo. The reel's rating keys
are all photo:<id> now, and the QR card's filter follows the sunset preset it
claims rather than an index that moved.
The claim was printed in four places — the RAW feature card, the quality FAQ,
the custom-recipe readout and the Lite plan list. Each now reads as
print-ready instead, and the readout keeps only RECIPE CUSTOM_01 · EXIF KEPT.
The exporter's own JFIF density is untouched, web-smoke still checks it.
The row of pills grew as the library did, so the landing's live tester now
offers the stocks through one native select, grouped into Landscape, Portrait
and Streetlife, five stocks each — fifteen in all. Choosing one still lands on
the preview immediately: the frame, the HUD and the spec card all follow the
selection. Stock names stay proper nouns, the group names are translated with
the rest of the page.
The card hangs on the point the eyedropper read, which is exactly where the
user wants to watch the band move — so it covers the patch it is editing. Its
body now takes a drag: the offset is a fraction of the photo, which is the
layer the card lives in, so a zoom keeps it where it was put and a fresh pick
drops it back on its own point. The knobs and the close button keep the pointer
to themselves, and the anchor cannot leave the photo, so the card is always
half in reach of a drag back.
The split used to paint the photo file beside the render, so it only lined up
while nothing had moved: a turn, a straighten, a printed frame and the halves
were two different pictures. The app now renders the same frame twice — once
through the look, once through the neutral stock — and the left of the bar is
that second copy. Rotation, straighten, crop and frame land on both halves by
construction, so the CSS that tried to map the crop onto the file goes away.
The toggle lives in the app now, which is what knows how to ask for the extra
render; it is only asked for while the split is up. The layer waits for that
copy rather than flashing the raw file, whose geometry is already wrong.
Choosing a crop ratio used to disable COMPARE outright, because the split
painted the whole original into a box that was now the crop's shape. The
original is now looked at through a window of the render's own shape: with a
crop applied the photo is scaled and slid by the crop rect so the same
rectangle lines up, and the split compares like with like.
The window needs the image free to overflow it, so the inline style lifts the
box clamp that .canvas-wrap img puts on every preview.
CROP + APPLY then a wall frame handed back the whole photo: the crop block
was skipped outright for both walls, so the artwork hung the original. The
walls' own opening still ignores the aspect chip — that is what the
exclusion was for — but the visitor's crop is theirs to keep.
Measured with a source banded red on top and blue below, cut away by a
16:9 crop: through WALL FRAME and WALL FRAME LANDSCAPE the bands used to
come back (569k and 350k red pixels); both now export clean.
Undo and redo were on the far side of the spacer, past RESET, SAVE PHOTO
and EXPORT. They belong next to what they step through: the header now
reads mark, page name, preset name, the two arrows, then the actions.
A tile's destination rectangle was placed at x0 * scale, and that scale
is rarely whole, so every 256px boundary landed on a fraction of a pixel.
The edge was drawn half covered, stayed transparent, and the JPEG export
flattened that transparency onto black: a dark line down each seam.
Snap both destination edges to whole pixels instead, so neighbouring
tiles share the exact same boundary, and make the destination context
opaque so no partly covered pixel can survive as transparency again.
Measured on a 640px source: the seam at 2K was 46 levels darker than its
neighbours (96 at 4K); it is now within one level of them.
The server still never sees a photo, so the model has to run in the page.
Real-ESRGAN x4v3 ships as a 4.9MB ONNX in public/models and is loaded
lazily on the first export that actually needs it; the wasm runtime is
copied next to CanvasKit at build time and stays lazily fetched, cached
for 30 days. Vite is told onnxruntime-web is external-wasm so no 28MB
asset lands in the bundle.
UNCHANGED keeps the old path and the tier cap; 2K/4K/custom upscale only
when the request is larger than the photo being edited, otherwise they
resize down. Guests keep UNCHANGED and 2K. Tiling is 256px with an 8px
overlap, so memory follows the target size rather than four times it.
The histogram used to park itself in the top-right corner on the first
paint; it now starts at the top-left of the photo and is dragged from
there, the way the rest of the overlay is. Nothing else changed in it —
same drag, same clamping, same resize.
The stage also gains a CLEAR button, sitting before the picker button,
which is now OPEN PHOTO. CLEAR takes the photo off the stage, but not
before asking: SAVE PHOTO files it first and only then clears, EXPORT
IMAGE writes the JPEG and then clears, CLEAR WITHOUT SAVING drops it
there and then, and CANCEL leaves everything alone. Saving from that
modal resumes the clear once the file has really landed — a guest, a
capped account or a cancelled name prompt never loses the frame.
Clearing forgets the working photo (source, preview, GPS, ISO, and the
IndexedDB copy session.ts now deletes), while the look, the crop and the
undo history stay put, so the next photo opens on the same settings the
way replacing a photo already did.
The name table is a reference for what each sim has to look like, not a
renaming order: the ten PHOTO STYLE chips go back to PROVIPES, VELVIPES,
CLASSIC CHRIPES, CLASSIC VIVIDIPES, CLASSIC NEGIPES, ASTIPES, ETERNIPES,
ACRIPES, LC STREETLIFE CLASSIC and LC STREETLIFE VIVID. The comment above
FILM_SIMS now says so outright — label on the left, the stock's colour and tone
it must match on the right, and neither side moves the other.
The grading is untouched: a sim is still colour and tone only, its `adjustments`
stay neutral, and LC STREETLIFE VIVID keeps its +2 exposure as SIM_EXPOSURE_BIAS
in colorUtils rather than as a knob.
The ten PHOTO STYLE sims now carry nothing but their stock's own grade, and
each is named for the stock it stands for: PROVIA, VELVIA, CLASSIC CHROME,
CLASSIC VIVID (Velvia spliced with Classic Chrome at the blue row), CLASSIC
NEGATIVE, ASTIA, ETERNA, ACROS, LC STREETLIFE CLASSIC, LC STREETLIFE VIVID.
Grain, clarity, saturation and light moves were dropped from their
`adjustments`, so a sim is a clean starting point and the general knobs read
their defaults while the look still lands on the pixels.
LC STREETLIFE VIVID keeps the one brightness step its stock needs, but as
SIM_EXPOSURE_BIAS in colorUtils rather than as an adjustment: it is folded in
where the Exposure slider applies, so the picture gets the lift and the
parameter stays at 0.
Also in this checkpoint: the watermark/GPS boxes and their colour pickers, the
WATERMARK chip column, the real admin stats, and the fix that stopped presets
from doubling and a frame from refusing to come off when a photo was reopened
(/file is the finished render, /base the editable pixels).
A signed-in account is served exactly like a guest until it opens the
verification link: watermarked 2048px export, no saving, no PRO frames,
GPS stamp or HDF. SMTP is declared in .env; with SMTP_HOST unset the link
goes to the container log. Allowlisted admins count as verified.
Tapping a tab on a phone opened a 148px column with the chips stacked one per
line, so the strip read as a ladder down the side of the photo. Android's own
panel runs its chips as a row — see src/components/AdjustmentPanel.tsx — and
that is what the phone now gets: the columns stack into one vertical scroll
and each chip row runs sideways again, wrapping inside the full width.
Desktop and tablet keep the columns and the stacked chips; the change lives in
the <=860px block.
On a phone the studio's tabs are now the row the Android app draws: text
pills in uppercase mono, rounded full, amber and a step larger when open, no
glyph, scrolling sideways when the ten tabs outrun the screen.
Desktop keeps its icon-over-label column — the change lives in the <=860px
block, so nothing above that breakpoint moves.
A session followed the browser, not the account: log in, open a frame, log
out, come back as a guest — the same photo stood on the stage, because the
studio's own store (localStorage knobs + the photo in IndexedDB) outlived the
cookie with nothing to clear it.
clearSession() now drops both, and the three log-out buttons call it. The
studio's own button reloads after the delete has committed — a reload mid-
delete aborts the transaction, so the promise resolves on tx.oncomplete, not
on the request. The account's frames are untouched: they reopen from MY
PHOTOS.
Both columns now have the shape the curator asked for: a narrow shelf of
albums down the left — one per account on the uploads side, one per landing
section on the other — the frame that is up in the middle, and the open
album's thumbnails as a strip across the bottom.
The frame keeps its labels, the four section boxes, the look's QR code and
the delete button under the picture, where before they sat beside it. Each
column previews its own frame; both obey the same name, sort and rating
filters.
The pane is two columns of the same thing: the uploads on the left, one
album per account, and the landing on the right, one album per section —
Film strip, Live preset tester, Custom recipe creator. QR is no longer a
shelf of its own: the code belongs to the frame.
Both columns list their albums the same way and draw the open album's
frames as cards, each with its labels, the four section boxes, its QR
code and the delete button under the picture. Ticking a box files the
frame into that album on the other side straight away.
The landing strip now carries a score: each look and each contributed
frame shows an average, five stars the visitor can press, and how many
votes it has. Votes are keyed photo:<id> or look:<TAG> and one visitor
has one vote per key, so pressing a second star moves a score instead of
stacking one. The API is public and rate-limited; look: scores survive a
cleared pool, photo: scores are pruned with their photo.
PICTURES was four destination rows; it is now an album per uploader with
a search box, a recipe/rating/newest sort, a minimum-star filter, a big
preview and a filmstrip of thumbnails. Deleting and slot picking still
live in the big box.
HUE, SAT and LUM leave the colour row: behind an IMAGE divider they are
hslHue/hslSat/hslLum, seeded into the shader's band accumulator at full
weight for every hue, while the eight band chips keep picking which colour
the panel on the photo edits. The image lightness term stays ungated so a
frame drained to grey by -SAT still answers +LUM.
FRAME loses its NO FRAME chip: pressing the frame already on the photo
takes it off.
ROTATE's quarter turns stop lighting the moment the fine angle leaves 0,
so the strip shows which of the two is steering the photo.
A photo's landing section can now be the QR card, and that section is the
only one that hands something out: the server writes the photo's own stored
look back as the app's .recipe file, at
GET /api/photos/:id/preset.recipe, for any row the curator ticked into the
qr slot. Nothing new is stored — the file is built from the recipe the
upload already carried, so it works for a photo uploaded by the phone too.
The admin pane grows a fourth checkbox and a fourth row (QR card); the
row draws the download link as a scannable code, and the box is dead for a
photo with no stored look. The landing's QR card now encodes the curated
photo's own link instead of a mock address. The listing exposes
hasPreset, never the recipe itself.
A save cannot be beaten — the bytes are already on the machine — so the page
stops handing over the uploader's 4000px original: each photo is decoded,
redrawn at the size of the box it sits in times the screen's pixel ratio (2x at
most) and only that smaller copy reaches the tag. A visitor who saves one gets
a screen-sized file.
Right-click, drag and long-press are turned off on top of it, and the QR code
card — a link image, not a contribution — is left as it was.
The picker was one dropdown, so a photo lived in exactly one place. The three
destinations are now independent checkboxes on the card, and the column holds
the set as a comma list — the landing page draws a photo in every section it
was ticked into, each still picking one of its own at random per visit.
Ticking nothing is what `off` used to be: the row is kept and the landing page
stops drawing it, which is what the old "not on the landing page" option did.
The pictures pane now reads as a pool of every upload on the left, three
destination rows on the right. Each photo carries one button per
destination; picking the one it already sits in takes it off the landing
page without deleting the row (slot off), which is what the old select's
"not on the landing page" option did.
3dtours