Repository navigation
feat(rigboy): v2 parts cut at every joint, with the ankle bend within bar - #87
Merged
Merged
Conversation
Five new region slots carry exact RGBA crops of neighbouring textures beneath the groin, rear knee, front ankle and rear wrist. Walk (31 samples, 16 pairs): open cells on those seams fall from 8/23/3/10 to 0, ghost is unchanged on every row, and the gate stays 17 pass, 0 fail. Bones, constraints and animations are identical to the accepted skeleton. Rejected: full-disc pieces (expose up to 267 px per frame on background) and a shoulder piece (the open cells are plain background, so it would be a phantom arm). Not clean by eye: the groin and front-ankle pieces show as small scraps at some frames, and no bar measures that. Co-Authored-By: Roa <342598975+fj-roa[bot]@users.noreply.github.com>
Hip and shoulder joints sat 15.59 to 37.87 px off the art. Five bones' setup positions move, and all four joints now measure 0.00 px from it. Rotations, constraints and animations are identical. Four under-pieces replace the round 1 set. Each donor was chosen so its exposed texels hold no orange or pink; the groin scrap and ankle fleck are gone and no ankle piece is needed. 499 texels of pink See-through filler on the rear arm take the arm's own texels. Walk, 31 samples: the four seams are at 0, but the moved rear hip opens torso:rear-thigh by 3 px at two frames, so that bar fails. The one stated repair was refused: rear-thigh has no art at those cells and the other direction lands on belt orange. Ghost is worse than the old rig on 10 rows and better on 4; the report gives each row's reason. Co-Authored-By: Roa <342598975+fj-roa[bot]@users.noreply.github.com>
…pieces Cut limbs stay; the piece that goes under at each joint gets a drawn round end. Code fill beat masked inpaint at one elbow and one knee: inpaint's B4 was .27 to .69 before palette snap against a .95 bar, and it equalled code after. Twelve limb joints now read this way, each part on one bone, no mesh across a joint, and the four under-* slots are gone. Bones, constraints, animations and events are identical to round 2. Compile gate: 17 pass, 0 fail, 19 skip, 14 not in profile. Ankle slots swap because the gold cuff is painted over the shin, so six slots moved in draw order. Walk, 31 samples: the gap bar fails on 6 of 16 rows, 1007 px in all, four of them worse than round 2. The cells are background beside the limbs, at least 11.0 px from a pivot. Joint discs read 0 at 10 of 12 joints. Rest against the painting is 7768 px, up from 7721. Co-Authored-By: Roa <342598975+fj-roa[bot]@users.noreply.github.com>
The seam reader's boxes sat on background, so the bar is now a joint disc of half the limb's fitted width. First read: 5 of 12 joints above 0, all at the rim. The cap's disc now sits on the pivot and reaches half the width + .25 + the joint's registration offset (up to 1.00 px at the near knee). Final read, 31 frames: 0 at 11 of 12 joints. The far shoulder reads 14 frames, 37 px, at the torso's back edge; the painted silhouette reads 14 px in that disc, so it stays. Stripping the drawn end reads 1161 px at the knee and 250 px at the elbow, so the gate fires. The orange ticks were the band fill's: it read the over piece's gold tail and laid a warm stripe, whose edge showed once the joint bent. A fill no longer invents gold (213 px), and warm components under 30 px are residue (50 px). Rest against the painting rises to 7824 px from round 2's 7721. Co-Authored-By: Roa <342598975+fj-roa[bot]@users.noreply.github.com>
gun.png was 730x400 at attachment scale 0.167 with no mipmaps, so it aliased into hard dither in a walk where every other region is 1:1. Any region drawn under 0.5 scale now goes to its drawn size, area-averaged on premultiplied pixels; the gun is the only one, at 122x67. Width, height and scale stay in the spec, so world size holds and the skeleton's md5 is unchanged. The joint disc now takes the painting as its ceiling: R is the lesser of half the limb's width and the pivot's distance to the painted edge at rest, which moves four joints. It reads 0 at all 12 on 31 frames, and stripping the drawn end still reads 1141 px at the knee and 250 px at the elbow. Co-Authored-By: Roa <342598975+fj-roa[bot]@users.noreply.github.com>
…source The leg is 134 long on a hip 142 high, against 204 on 247 in the source, so the same foot path bent the knee further and laid the shin over. Round 6 changes foot-path geometry only: hip y +7, front-foot-target x +30, and the x of both walk foot targets' translate keys times 0.85. No bone length moved and the images are round 5's, byte for byte. Walk, 31 samples, shin-to-foot bend against the source: front 13.8 degrees at worst (35.6 before), rear 12.0 (21.1 before), samples past 15 degrees 0 of 62 (28 before). 0 background pixels at all 12 joint discs. The cost: stride and ground speed drop 15 %, and the feet are 80 px apart at one double support, where they were 124. Rejected: chasing the rigid boot's remaining over-rotation at heel-off, which is inside the bar. Round 5 stays out of the tree: it was superseded before landing. Refs #80 Co-Authored-By: Roa <342598975+fj-roa[bot]@users.noreply.github.com>
Codex Review SummaryThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 945d0381e9
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| "images": "./images/", | ||
| "audio": null | ||
| }, | ||
| "images": "../images/", |
There was a problem hiding this comment.
In a clean checkout, this rig points to rigboy/v2/images/, which is not included, while the new limb PNGs are stored alongside rig.json. Changing the image root alone is insufficient: the default skin also requires eye.png, goggles.png, head.png, mouth.png, and neck.png, which are absent from this round. I checked the repository-wide tracked-file list, including the existing rigboy/build artifacts; there is no corresponding v2 image bundle or packed fallback. Consequently, the published round cannot be loaded or packed as a complete character without unpublished inputs. Include the remaining textures (or a reproducible acquisition step) and point the rig at the supplied image directory; the other three new rig specs have the same dependency problem.
Useful? React with 👍 / 👎.
…checkout Round 6's rig reads ../images/, which was not in the tree, and five parts the default skin needs (eye, goggles, head, mouth, neck) were in no tracked round. Move the round's limb PNGs to rigboy/v2/images/ with those five. All 20 images the skin names resolve, none is unused, the 13 limb pieces match round 5's sealed hashes, and 0 match a Spineboy PNG by SHA-256. rigc build from the tree alone exits 0 with core 2.84.0. Refs #80 Co-Authored-By: Roa <342598975+fj-roa[bot]@users.noreply.github.com>
Collaborator
Author
|
P2 on the v2 image dependencies: adopted, it was a real defect. A clean checkout could not load the rig. Fixed in this PR by carrying the full image set in |
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.
The leg is 134 long on a hip 142 high, against 204 on 247 in the source,
so the same foot path bent the knee further and laid the shin over.
Round 6 changes foot-path geometry only: hip y +7, front-foot-target x +30,
and the x of both walk foot targets' translate keys times 0.85. No bone
length moved and the images are round 5's, byte for byte.
Walk, 31 samples, shin-to-foot bend against the source: front 13.8 degrees
at worst (35.6 before), rear 12.0 (21.1 before), samples past 15 degrees
0 of 62 (28 before). 0 background pixels at all 12 joint discs.
The cost: stride and ground speed drop 15 %, and the feet are 80 px apart at
one double support, where they were 124. Rejected: chasing the rigid boot's
remaining over-rotation at heel-off, which is inside the bar.
Round 5 stays out of the tree: it was superseded before landing.
This branch carries the v2 parts, cut at every joint, up to round 6.
rigboy/v2/cut-ends-round6/is the current state.No Spineboy image in the tree: 0 of the round's PNGs match the 42 Spineboy PNGs by SHA-256. Comparison media stay private.
Refs #80
🤖 Generated with Claude Code