Repository navigation
Two canvases cloned from screen recordings - #179
Open
Jing-yilin wants to merge 3 commits into
Open
Jing-yilin wants to merge 3 commits into
Jing-yilin wants to merge 3 commits into
Conversation
Seven boards rebuilt from an 8.2 s, 2560 x 1698 recording of an unnamed events-feed concept: a token board, two evidence boards for 48 tokens, and four screens that are one scrolling feed at four scroll offsets. The source is a video rather than screenshots, which is new here and is what most of the folder's README is about. No frame holds the whole screen sharply, because the camera zooms and pans over a phone that never moves, so the run solves the screen's edges per frame, finds three static camera setups at 1.45, 3.61 and 3.78 px/pt, and probes.json carries all three with each probe naming its own scale. A frame is also not a screenshot: two frames of the same static shot differ by mean 1.14 to 1.69 levels, and that is the floor the deltas are read against. Mean absolute delta against the captures is 7.29 to 11.56 levels (of 255) with the photographic regions masked and 13.22 to 20.96 with them in. The photo stacks and the avatars are original gradient art drawn from each region's measured colours, because the source's photographs belong to whoever made the recording; scratch/exart.py is what separates the two numbers. The phone is a 440 x 956 iPhone 17 Pro Max, which does not fit the default 478 x 980 artboard, so every board declares 532 x 1004 in layout.json. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A region file is written by hand and reused across a run, so it wants a note saying what each box is and why it is drawn where it is. A probes.json already spells that convention: a key starting with an underscore is a comment. The diff command did not, and unpacked the note as four coordinates, so the table crashed halfway through and the run silently lost its per-region numbers. Skip those keys, say so in the --regions help and in the skill, and add the regression test. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Five boards plus two reference boards from a 9.9 s, 2160 x 2160 recording of an iOS app called Stepwise: a token board and two evidence boards for 43 tokens, then the profile screen and the "Pick your vibe" avatar sheet over it. Two states of the recording hold still long enough to measure, and the camera zooms in at the same moment the sheet animates, so the two references sit at different scales and each probe carries its own. Mean absolute delta against those frames is 3.38 and 1.51 levels (of 255) once the substituted regions are cut, 7.24 and 21.27 whole. Every face on these boards is this repo's own character art: the source's illustrations are its designer's and are not reproduced, traced or imitated, so what was measured and matched is the system around them, the 2.78 pt line on a 92.7 pt disc, the ink box, the pale ground and the grid pitch. Those regions cost 72.54 and 41.65 and the README states them rather than folding them into a total. Also: av-11's hair path was malformed and rendered as a stray wedge across the face, redrawn as a fringe in the same style; and events-feed-ios drops the `order` it inherited from the templates folder, so both new examples sort with the other clones rather than below the utility folders. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
Deploying super-prototyping with
|
| Latest commit: |
97ff8f5
|
| Status: | ✅ Deploy successful! |
| Preview URL: | https://3399f64f.super-prototyping.pages.dev |
| Branch Preview URL: | https://clone-video-prototype.super-prototyping.pages.dev |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Two new example folders, both cloned from short screen recordings rather than
from screenshots or a Figma file. That source is new here and it is most of
what the two READMEs are about: no frame of a recording holds a whole screen
at full resolution, a frame is not a screenshot so the delta floor is not
zero, and h.264 blocking sits on every rim.
canvases/events-feed-ios/— 7 boards + 4 referencesAn unnamed events-feed concept, 8.2 s at 2560 x 1698: a token board, two
evidence boards for 48 tokens, and four screens that are one scrolling feed at
four scroll offsets.
The camera zooms and pans over a phone that never moves, so the run solves the
screen's edges per frame, finds three static camera setups at 1.45, 3.61 and
3.78 px/pt, and
probes.jsoncarries all three with each probe naming its ownscale. Two frames of the same static shot differ by mean 1.14-1.69 levels, and
that is the floor the numbers are read against.
Colour probes 11, mean Δmax 1.3. Box probes 23, mean |dw| 0.82, mean |dh| 0.24.
The photo stacks and the avatars are original gradient art drawn from each
region's measured colours, because the source's photographs belong to whoever
made the recording;
scratch/exart.pyis what separates the two columns. Thephone is a 440 x 956 iPhone 17 Pro Max, which does not fit the default
478 x 980 artboard, so every board declares 532 x 1004.
canvases/stepwise-ios/— 5 boards + 2 referencesAn iOS app called Stepwise, 9.9 s at 2160 x 2160: a token board, two evidence
boards for 43 tokens, the profile screen and the "Pick your vibe" avatar sheet
over it.
Colour probes 18, mean Δmax 2.6. Box probes 34, mean |dw| 0.50, mean |dh| 0.24.
Every face is this repo's own character art. The source's illustrations are
its designer's and are not reproduced, traced, cropped or imitated; what was
measured and matched is the system around them — a 2.78 pt line on a 92.7 pt
disc, the ink box, the pale ground, the grid pitch — and the two art regions
cost 72.54 and 41.65, stated rather than folded into a total. The capture
scale does not close to itself (2.38185 px/pt across against 2.36819 down, on
a frame whose 2.157 aspect belongs to no shipping iPhone), and the top
69.93 pt of the sheet screen is simply not in the recording.
One tool fix
refkit diff --regionscrashed on a region file carrying a_-prefixed note,unpacking the note as four coordinates and losing the rest of the table. A
probes.json already treats such a key as a comment; the diff command now does
too, with the help text, the skill and a regression test to match.
Notes
ref-*.htmland noassets/refs/are committed; the root.gitignorealready excludes both, so a fresh clone gets 7 boards and 5.
python3 canvases/<slug>/gen.py, passrefkit tokens, and fit theirdeclared artboard under
--check-overflow.tools/test_refkit.pypasses 26/26 under bothpytestandpython3 tools/test_refkit.py.🤖 Generated with Claude Code