Skip to content

Commit c5bc55b

Browse files
committed
OOO/OBR: build alpha92 devlog entry (6 new sets + self-verifying gates)
Add the alpha92 section to docs/OOO/OBR/devlog.html above alpha91: intro, five arc cards (Daedric family, Dragonborne emblem, recolor-pipeline extensions, self-verifying builds, Draconis) and a 12-milestone spine spanning 2026-06-14..06-27. Six set renders hosted in images/refs/ (render-only milestones; piece icons deferred to the new-Items alpha92 update). Draconis framed as confirmed in-game per user retest. alpha92 is in progress/unreleased; entry covers work so far.
1 parent f99f73a commit c5bc55b

8 files changed

Lines changed: 103 additions & 4 deletions

docs/OOO/OBR/devlog.html

Lines changed: 99 additions & 0 deletions
Large diffs are not rendered by default.
2.27 MB
Loading
3.58 MB
Loading
3.51 MB
Loading
2.44 MB
Loading
2.5 MB
Loading
2.44 MB
Loading

notes/devlog-alpha92-draft.md

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -10,15 +10,15 @@ alpha92 is in progress. The map work is settled as of alpha91, so the work so fa
1010
- **The Dragonborne emblem** | `a92-emblem` | Replacing the Kvatch wolf with a dragon insignia on the cuirass took four days. The emblem was traced to the wrong layer twice before it was found in the cloth's normal map, hidden by a quirk of how the export tool drops channels.
1111
- **Extending the recolor pipeline** | `a92-pipeline` | The new sets needed capabilities the older recolors did not have: passing the vanilla diffuse through untouched and editing only named regions, shipping a hand-painted texture as-is, and overriding only part of a texture while holding the rest vanilla.
1212
- **Self-verifying builds** | `a92-gates` | The pipeline now turns each config promise into a build-time check that inspects the output and refuses to emit on a violation. A capstone gate requires every output-shaping directive to carry its own verifier, so a new knob cannot be added without a check.
13-
- **Draconis** | `a92-draconis` | A two-piece Kvatch derivative with twin red griffins, whose plate refused to take a steel color until the cause was traced past the texture to a blue tint baked into the shared parent material.
13+
- **Draconis** | `a92-draconis` | A two-piece Kvatch derivative with twin red griffins, whose shoulder plates refused to take a steel color until the cause was traced past the texture to a tint baked into the shared parent material.
1414

1515
## Chronological spine (oldest -> newest, 2026-06-14 -> 2026-06-27)
1616

1717
- `2026-06-14 | - | Dragonborne scoped and built through clone | Dragonborne is a five-piece dark recolor of the Kvatch/Townguard "Townguard" armor with no helmet: a near-black quilted tunic, cooled steel, and dark-brown leather, with the Kvatch wolf emblem to be replaced by a dragon insignia. Reconnaissance traced the full asset chain and surfaced the cuirass as the hard piece — its body mesh and cloth-simulation assets come from one vanilla folder while its material, textures, and ground mesh come from another, with inconsistent casing between them. The set was cloned and built end-to-end today, rendering as a dark Townguard set with the wolf emblem still in place. [THUMB: Dragonborne]`
1818
- `2026-06-16 | a92-emblem | The Dragonborne emblem: four days on the wrong layer | Replacing the wolf with a dragon spanned 06-14 to 06-18. On the cuirass the dragon is a flat painted decal; on the shield it is an embossed metal relief. The first obstacle was a notch where the cuirass emblem's frame met its picture: the frame's bottom verts were sampling the wrong texels of the disc, which was fixed by fitting the frame's mapping only from the clean rim ring. The deeper trap came next. The old wolf kept showing through, and was diagnosed as baked into the mesh — a wipe of the mesh did nothing in-game. The real cause was that the export tool drops a packed texture's color channels wherever its metallic channel reads zero, which is the entire cloth region, so the emblem only looked absent. The wolf was actually in the cloth's normal map; flattening it there, then adding back clean weave grain, removed it. This was confirmed in-game.`
1919
- `2026-06-18 | - | Dragonborne cuirass trim and shield rivets | Two finishing issues closed out the cuirass. The skirt and strap leather read as plain brown rather than the set's oxblood, because the tunic's cloth mask suppressed the path that gave the other pieces their red; a cuirass-specific recipe restored it. On the body pieces the warm recesses of the chainmail were being painted red as if they were cloth, fixed by gating the cloth recoloring to non-metallic pixels so the mail reclaims its steel. The shield's rivets were over-grown by an unnecessary dilation, which was removed. A thin trim around the tunic panels reading three different colors was noted as a remaining polish item.`
2020
- `2026-06-19 | a92-draconis | Draconis built; passthrough diffuse and region-class recoloring | Draconis is a two-piece Kvatch derivative — a cuirass and shield carrying twin red griffins in place of the wolf. It needed an approach the earlier global recolors could not provide, since the vanilla armor was to be preserved and only specific patches changed. Three mechanisms were built for this: passing the vanilla diffuse through untouched and editing only named regions; giving a region its own surface recipe so the shoulder plates and shield boss could be flipped to metal while the rest stayed vanilla; and subtracting one region from another, used to keep the painted griffin out of the metal boss. The set was built and confirmed in-game the next day, which exposed the bugs handled below. [THUMB: Draconis]`
21-
- `2026-06-20 | - | Draconis steel and the baked master tint | The converted plate regions rendered shiny brown on the cuirass and shiny black on the shield rather than steel, because on a metallic surface the base color tints the reflection, and the recipe had been derived from the dark leather beneath. Neutralizing to true steel — bright on the shield, darker gunmetal on the cuirass — addressed the shield, but the cuirass plates stayed near-black no matter the texture. The cause was not in the texture at all: the cuirass material inherited a shared parent that bakes in a strong blue tint at full strength, which crushes any neutral or dark base toward near-black. Repointing the cuirass to a plain untinted parent let the steel render true. The clue was an overlay showing the visible plate faces sampling the edited texture by only about five percent, which pointed past the texture to the material.`
21+
- `2026-06-20 | - | Draconis steel and the baked master tint | The converted plate regions rendered shiny brown on the cuirass and shiny black on the shield rather than steel, because on a metallic surface the base color tints the reflection, and the recipe had been derived from the dark leather beneath. Neutralizing to true steel — bright on the shield, darker gunmetal on the cuirass — addressed the shield, but the cuirass plates stayed near-black no matter the texture. The cause was not in the texture at all: the cuirass material inherited a shared parent that bakes in a strong blue tint at full strength, which crushes any neutral or dark base toward near-black. Repointing the cuirass to a plain untinted parent let the steel render true. The clue was an overlay showing the visible plate faces sampling the edited texture by only about five percent, which pointed past the texture to the material. With both pieces corrected, Draconis was confirmed in-game and called done.`
2222
- `2026-06-20 | - | The Daedric base and Dread | The Daedric family is four sets cloned from one shared Daedric body, so a fix to the shared cuirass propagates to every variant. Reconnaissance found the Daedric diffuse textures use a format with a load-bearing alpha channel, not the simpler format the plan assumed, and corrected a decoder that had been misreading them. Dread, the first set, is a dark green-steel plate with a blue gem. Its first build still showed vanilla red in-game on the trim, because part of the red lives in the material's color parameters rather than the visible texture; a tool that rotates those parameters to the target hue resolved it. The skirt's dark-red cloth hem was too unsaturated for the automatic mask to catch and was handled with a hand-painted region. [THUMB: Dread]`
2323
- `2026-06-22 | - | Dread metal and skirt panels | The green metal recolor had flattened the texture's brightness range so the mesh relief stopped reading; a value-preserving recolor following a supplied olive swatch restored it. The same diagnosis confirmed and fixed the texture-format decoder bug. The side skirt panels — fragmented into many small separately-mapped scales — were left showing vanilla brown, which fed into the later shared side-panel work. Dread ships a small hand-edited diffuse to clean up the remaining green bleed.`
2424
- `2026-06-23 | - | DaedricPrince: lava, and where the glow comes from | DaedricPrince is a six-piece lava set, including a shield: tarnished bronze-gold metal that glows with red lava, indigo chainmail, and a gem rendered as a radial flame burst. Two techniques were built for it — a value-dependent hue ramp that reads each pixel's brightness so dark areas stay deep red while highlights shift to orange, and a radial fill that brightens the gem at its center and darkens it to the rim. The first build rotated the glow's color parameters toward orange, which killed the vibrant red on every piece except the cuirass. The cuirass survived because its glow comes from a rich texture alpha mask, not the parameters; the shield, having no such mask, went dull. The fix set the glow parameters back to pure red and let the diffuse supply the warmth. [THUMB: DaedricPrince]`
@@ -30,8 +30,8 @@ alpha92 is in progress. The map work is settled as of alpha91, so the work so fa
3030
## Fact-check notes
3131

3232
- **Dates are author dates from git where set milestones are concerned.** Dragonborne recon/clone and the emblem fix span 06-14 to 06-18 (merged 06-18); Draconis built 06-19, fixed 06-20; the Daedric base and Dread 06-20 to 06-22; DaedricPrince 06-23; shared eyes/gem fixes 06-24; ImagoStorm and DremoraElite 06-25; Daedric icon-aura rebuild 06-26; the config-honored gate framework 06-27. Where a handoff filename's date differs from the commit date, the commit date is used.
33-
- **Set status.** All four Daedric sets and Draconis were confirmed in-game within this window (Dread and DaedricPrince on 06-24; ImagoStorm and DremoraElite by 06-27 after the deployment fix). Dragonborne's emblem and cuirass were confirmed in-game during the 06-14→06-18 push, and the set was rebuilt 06-27 with the dragon verified by the new gate. The Draconis handoff records its final fixed version as built and gate-green but pending one in-game retest at 06-20; treat its in-game status as confirmed-with-a-retest-pending unless a later in-game pass is on record.
33+
- **Set status.** All four Daedric sets and Draconis were confirmed in-game within this window (Dread and DaedricPrince on 06-24; ImagoStorm and DremoraElite by 06-27 after the deployment fix). Dragonborne's emblem and cuirass were confirmed in-game during the 06-14→06-18 push, and the set was rebuilt 06-27 with the dragon verified by the new gate. Draconis's final fixed version was confirmed in-game (user-verified after the 06-20 fix).
3434
- **In-progress framing.** alpha92 is not released. Nothing here is described as shipped; each item is "built" or "confirmed in-game," and the entry covers work so far.
3535
- **Open items carried, not resolved.** The Dragonborne tunic trim reading three colors was an open polish item at 06-18. One documented Daedric mystery — ImagoStorm's greaves showing more emissive swirl than its cuirass despite identical material data — was de-prioritized and accepted, not root-caused. Neither blocks the sets.
36-
- **Could not independently verify.** Exact in-game retest confirmation for the final Draconis version after 06-20 was not located in the read sources; the gate-green build is confirmed. The four-day emblem gap is genuine (06-14 build, emblem resolved 06-17→06-18) and is reported as a real debugging span, not an attempt count.
36+
- **Could not independently verify.** The four-day emblem gap is genuine (06-14 build, emblem resolved 06-17→06-18) and is reported as a real debugging span, not an attempt count.
3737
- **No raw identifiers in the prose.** Asset names, FormID hex, version tags, and commit hashes from the sources are translated out of the entry.

0 commit comments

Comments
 (0)