docs: visual asset options menu + pipeline inventory (Pip pick-and-iterate) - #649
Merged
Conversation
…erate) Menu-of-dials decision doc for rapid visual iteration alongside playtesting. Inventories the gpt-image-1.5 pipeline (runs today; only OPENAI_API_KEY missing), current asset coverage, and prior scoped ideas (hero_banners.yaml, ASSET_GENERATION_BATCH, ICON_MAPPING_AUDIT). Ranks gaps by player-visible impact and recommends a first batch (heroes + placeholder icons + portrait pilot, ~$2.85). No generation, no binary assets. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard. |
PipFoweraker
added a commit
that referenced
this pull request
Jul 21, 2026
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
PipFoweraker
added a commit
that referenced
this pull request
Jul 21, 2026
…649) Promote batch #649 gpt-image action icons (kills placeholder fallbacks): team_building v1, network v2, acquire_startup v2, sabotage_competitor v1, lobby_government v1 (stopgap -- re-roll queued), open_source_release v1, take_loan v1, grant_proposal v2. icon_mapping.json repointed; new test_icon_mapping.gd asserts all 8 resolve to real textures, not the magenta placeholder. Fast gate: 540 tests, 0 fail. Stage CRT-dossier v2 portrait set (5 archetypes, one coherent look) under godot/assets/portraits/ -- staged-but-unwired, no UI slot yet. README records the per-archetype variant-pool + random-assignment design. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
PipFoweraker
added a commit
that referenced
this pull request
Jul 22, 2026
gpt-image-1.5 re-rolls implementing each verdict re-roll note: - 8 action icons (2 variants, rendered OPAQUE to fix failing alpha edges) - 10 portrait entries (5 subjects x 2 varied demographics) for the per-archetype variant pool; fresh-grad aged to graduated-undergrad, less cartoon/anime/ghibli, ethnicity/gender varied - 3 banners (2 variants): title-hero (extra light + cat/comforts/poster), doom-rising (CONSERVATIVE in-palette only -- doom-colour spec owed), strategic-moves (chair separated/lit, shorter FOV/table) Manifests: art_prompts/round3_rerolls.yaml (+_banners.yaml, split for landscape). Review gallery rebuilt. NOTE: only the sub-1000KB downscales are committed (the 512/768 sizes the gallery shows); the 1024/1536 masters exceed the check-added-large-files guard and stay on disk (regenerable from the manifests). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
PipFoweraker
added a commit
that referenced
this pull request
Jul 22, 2026
… frames (#649 approved) (#761) * art(assets): first gpt-image-1.5 batch -- banners, icons, portrait A/B, CRT frames Fires the PR #649-approved visual-asset batch (docs/art/VISUAL_ASSET_OPTIONS_2026-07-16.md). Model id gpt-image-1.5 confirmed live via a throwaway probe (no fallback to gpt-image-1 needed). Batches (all via tools/assets/generate_images.py, YAML-as-source-of-truth): - hero_banners: 3 scoped banners x3 variants (9 renders). Added background: opaque so full-bleed skies/dark grounds are not alpha-cut (the generator defaults to transparent -- correct for icons, wrong for banners). - action_icons_missing (NEW yaml): 8 placeholder-killer icons x2 (16). Palette re-authored off the stale cyan/blue register to desaturated teal/olive + amber. - researcher_portraits_pilot (NEW yaml): 6 lore archetypes x 2 style poles x2 (24) -- painterly XCOM (pole A) vs flat CRT-dossier (pole B) A/B. Archetypes only, never real people (event-horizon guardrail). 6th archetype (fresh-grad true-believer) is a discretionary lore-grounded pick, mirror of the burned-out senior. - crt_frame_overlay (NEW yaml): 3 CRT bezel/frame mocks x1 (3), transparent centre. 52 renders, ~$3.57 estimated (per-image estimates; confirm true billing on the OpenAI dashboard -- the SDK image path does not surface token usage). Nothing promoted to godot/assets yet; Pip curates in the HTML gallery. Repo-hygiene notes: - art_generated/ is normally gitignored (curate-then-promote design); these were force-added by explicit path so the options ship in the PR for review. - The 52 master PNGs were palette-compressed (256-colour, alpha preserved) to clear the repo's 1000KB file guard. They are review-grade, not pixel-exact; the exact masters regenerate deterministically from the YAML prompt hashes. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * tools(assets): verdict UX in the asset gallery + 52-image batch merge Upgrades tools/assets/select_assets.py gallery generation to fold the tabbed-review verdict UX (tools/art_review/build.py) into the winner-picking asset gallery, so batch-649's 52 renders get the review flow Pip clears hundreds with: - K/M/R keyboard verdicts (keep/maybe/re-roll) on the focused cell, per-cell notes, status filter All/Pending/Keep/Maybe/Re-roll, hide-on-verdict (a decided cell leaves the Pending view), localStorage persistence, markdown export -- CSS/JS lifted from the tabbed viewer (dark cozy-grim palette). - Winner-pick per asset group kept: star a variant (W / click), copyable per-manifest --select id:vN commands. Click an image to enlarge; Esc closes. - Disk-based variant discovery replaces the size-[128..1024]-only path: now handles 768/1536 banners+CRT frames and the no-variant <id>_<size>.png CRT naming. Images referenced by RELATIVE path (portable), with an absolute file:// fallback only when output + images sit on different drives. HTML is written LF so it clears the mixed-line-ending hook. - New --gallery-batch <status> scans art_prompts/*.yaml and merges every asset with images on disk into ONE gallery (tabs per sub-batch + All). Single --file --gallery mode kept and routed through the same builder. Regenerated asset_gallery.html over the real batch: 52 cells across 4 tabs (action_icons_missing 16, hero_banners 9, researcher_portraits_pilot 24, crt_frame_overlay 3), all paths relative and resolving on disk. Force-added past the art-artifact gitignore so it renders in the PR for review. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * docs(art): batch #649 verdict record + portrait variant-pool ruling Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * art: promote 8 stopgap action icons + stage CRT-dossier portrait set (#649) Promote batch #649 gpt-image action icons (kills placeholder fallbacks): team_building v1, network v2, acquire_startup v2, sabotage_competitor v1, lobby_government v1 (stopgap -- re-roll queued), open_source_release v1, take_loan v1, grant_proposal v2. icon_mapping.json repointed; new test_icon_mapping.gd asserts all 8 resolve to real textures, not the magenta placeholder. Fast gate: 540 tests, 0 fail. Stage CRT-dossier v2 portrait set (5 archetypes, one coherent look) under godot/assets/portraits/ -- staged-but-unwired, no UI slot yet. README records the per-archetype variant-pool + random-assignment design. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> * art: round-3 re-rolls per #649 verdicts (32 images, $2.10) gpt-image-1.5 re-rolls implementing each verdict re-roll note: - 8 action icons (2 variants, rendered OPAQUE to fix failing alpha edges) - 10 portrait entries (5 subjects x 2 varied demographics) for the per-archetype variant pool; fresh-grad aged to graduated-undergrad, less cartoon/anime/ghibli, ethnicity/gender varied - 3 banners (2 variants): title-hero (extra light + cat/comforts/poster), doom-rising (CONSERVATIVE in-palette only -- doom-colour spec owed), strategic-moves (chair separated/lit, shorter FOV/table) Manifests: art_prompts/round3_rerolls.yaml (+_banners.yaml, split for landscape). Review gallery rebuilt. NOTE: only the sub-1000KB downscales are committed (the 512/768 sizes the gallery shows); the 1024/1536 masters exceed the check-added-large-files guard and stay on disk (regenerable from the manifests). Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> --------- Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
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
1 participant
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.
What
A menu of dials for visual assets so Pip can pick a direction and iterate fast in parallel with playtesting — plus a full pipeline + asset inventory. Research/prep lane: doc only. No images generated, no keys used, no binary assets committed. Do NOT merge — this is for Pip to pick from.
New file:
docs/art/VISUAL_ASSET_OPTIONS_2026-07-16.mdKey findings
tools/assets/generate_images.py(gpt-image-1.5 default; dormant Gemini "Nano Banana Pro" backend with reference-image consistency). Deps installed locally (openai 2.8.0, Pillow 12.0.0, pyyaml); dry-run verified. Only gap:OPENAI_API_KEYunset, and the default model id carries aTODO: confirm exact id vs live docs— do one throwaway image to confirm id + true cost before a paid batch.art_prompts/hero_banners.yaml(3 scene banners, already on-register, stillpending);docs/assets/ASSET_GENERATION_BATCH.md(126-asset list, but stale cyan/blue palette — needs refresh);ICON_MAPPING_AUDIT.md(8 icons still rendering as placeholders).Open decisions for Pip
Portrait style (flat CRT-dossier vs painterly XCOM) · matched cast (Gemini+ref) vs independent · confirm model id/cost · per-genre event banners now or later · CRT frame worth mocking. See §7.
🤖 Generated with Claude Code