💃 Free CC0 mannequin pose reference library for AI image tools (Nano Banana, ChatGPT, ComfyUI, Midjourney) — 250+ poses × 5 angles in line art, names in ja/zh/en/ko
Python ai-artai-image-generationart-referencecc0comfyuicontrolnetdatasetdrawing-referencefigure-drawingline-artmannequinmidjourneymultilingualnano-bananapose-referencereference-imagesstable-diffusion
docs: record the digi redraws (D-09, QC round 3) and update counts
wqugurud0527bd9f287
diff --git a/AGENTS.md b/AGENTS.mdindex 27e9a5ec..eaf7e4a0 100644--- a/AGENTS.md+++ b/AGENTS.md@@ -45,9 +45,10 @@ CONTRIBUTING.md contributor guide; SECURITY.md, CODE_OF_CONDUCT.md; .githubCHANGELOG.md Keep a Changelog; every user-visible change adds a line under [Unreleased]data/catalog.json 288 poses: id, orig_id, series, rating, added_in, class_*, name_* (jp/cn/en/kr), desc_en,views{front..below} (file or null), gaps{view: missing|line|structure}, photos{view: file},- bodies (duo only: MF / FF)+ bodies (duo only: MF / FF), source (redrawn sets only: digi, D-09)data/packs.json Release download zips: file, bytes, sets (tools/pack_release.py; the gallery shows the sizes)-data/qc.json per-view and per-photo QC status, reason and file name; per-set pose-name notes (docs/qc/README.md)+data/qc.json per-view and per-photo QC status, reason and file name; per-set pose-name notes; sources of+ redrawn sets (docs/qc/README.md)data/qc-input/ every reviewer result behind qc.json; `qc_merge.py data/qc-input` rebuilds it exactlydata/rating_overrides.json per-pose rating decisions that win over the category rule (human-decided)data/catalog.csv tools/export_csv.py output (o = passed, x = failed QC, - = not generated)@@ -62,7 +63,8 @@ site/ build output of tools/build_site.py (git-ignored; deploy th_headers (Cloudflare cache rules)build/og-cache/ share-card render cache (git-ignored)tools/ build_images.py, build_site.py (+ site_pages.py, og_cards.py), verify.py, pack_release.py,- import_series.py, review_sheets.py, line_dups.py, qc_merge.py, rating_check.py, export_csv.py+ import_series.py, import_redraws.py, review_sheets.py, line_dups.py, qc_merge.py, rating_check.py,+ export_csv.pydocs/qc/ QC method, results and the reviewer prompts (panels, name, limbs, proportion, photo, rating).claude/agents/ qc-reviewer: the Sonnet sub-agent every QC batch is sent todocs/DESIGN.md gallery layout, tokens, interaction spec (agent-facing copy of the design canvas)@@ -83,13 +85,15 @@ All series share `data/catalog.json`; a new series gets its own prefix and ID le## 5. Common commands```bashpython3 tools/build_images.py std=~/Downloads/pose-gallery-std73 sexy=~/Downloads/pose-gallery-sexy80 \- upper=~/Downloads/pose-gallery-upper80 duo=~/Downloads/pose-gallery-duo55 # passing views only; prunes the rest+ upper=~/Downloads/pose-gallery-upper80 duo=~/Downloads/pose-gallery-duo55 \+ digi=~/Downloads/pose-gallery-digi66 # passing views only; prunes the rest. digi = redrawn sets (D-09)python3 tools/verify.py # catalog ↔ qc.json ↔ images, gaps, orphanspython3 tools/import_series.py SRC --series X --added 0.2.0 [--general Cute] # merge a new series+python3 -I tools/import_redraws.py SRC OUT --source digi --only P019,… # whole-set redraws, then QC (D-09)python3 tools/qc_merge.py data/qc-input # apply review results (docs/qc/README.md)python3 tools/build_site.py --link-images && python3 -m http.server 8000 -d site # preview the gallerypython3 -I tools/build_site.py --public --site https://shisei.indaloop.app # deploy build (D-05); share cards need Chrome, --no-og skips-python3 -I tools/pack_release.py 0.2.0 std=… upper=… sexy=… duo=… # dist/*.zip + sha256, data/packs.json+python3 -I tools/pack_release.py 0.2.0 std=… upper=… sexy=… duo=… digi=… # dist/*.zip + sha256, data/packs.json```## 6. Way of working@@ -134,5 +138,6 @@ adding a new image series.| Fill P062 below | ☐ open || upper + sexy series, line-art QC, series / rating filters | ✅ v0.2.0 (2026-10-29) || duo series (55 sets) and photo QC | ✅ v0.2.0 (2026-10-29) |-| Relabel the 23 std sets whose drawing is another pose (D-07) | ☐ waiting on a human |+| Relabel the 23 std sets whose drawing is another pose (D-07) | ◐ 21 redrawn (D-09); P038, P052 open |+| Redraw gaps from a second source, whole sets only (D-09) | ✅ 43 sets, 2026-10-29 || Rewrite failed views out of git history before going public (D-08) | ✅ 2026-10-29 |diff --git a/CHANGELOG.md b/CHANGELOG.mdindex b50ffd26..7166731c 100644--- a/CHANGELOG.md+++ b/CHANGELOG.md@@ -8,12 +8,19 @@ the catalog schema.## [Unreleased]### Added+- 43 sets redrawn from a second generator (`source: digi`, D-09) where the original had at most two usable views or+ showed another pose: 21 of the 23 hidden std sets (D-07), U051 and U072 (never generated before) and 20 sparse+ sets. Whole sets only, after the full QC round 3. std now shows 71 of 73 sets, upper 80 of 80; 1,133 views pass.- Compare mode: the lightbox's source switch is now line art / photo / compare. Compare shows both at one height sideby side (no overlay: they are framed differently), keeps the gap box for a missing side, marks angles that have both,and offers both 900 downloads. `L` cycles the three; deep link `#P001-diag-compare`. Pose pages with photos get a「线稿与实拍对照」 section. The grid keeps two modes.### Changed+- The redrawn sets use a slightly different mannequin (construction lines, milder high and low angles), and their+ full-size originals are 900 × 1200.+- `qc_merge.py` scopes result files by `source`; `import_redraws.py` stages whole-set redraws; build, sheet, dup and+ pack tools take a `digi=` source; `verify.py` checks sources; `catalog.csv` gains a `source` column.- Phone layout is more compact. The filter sheet and its button are gone: the angle, source and view switches sit inthe sticky bar, which is down from 184 px to 114 px because the series row scrolls away. The grid defaults to threecolumns (about 11 poses per screen instead of 4.5), with a toggle back to two; the choice is kept in localStorage.@@ -40,6 +47,7 @@ the catalog schema.none instead of looking live and doing nothing.### Removed+- Photo renders of P019, P020 and P022 (14): they show the old drawing's pose, not the redrawn one.- P006 front view: the drawing has three arms, which the QC review missed. Manual QC results now go in`data/qc-input/fail_manual.json`.- Eight views a reviewer had failed but `qc_merge.py` let through: the floor-line rule matched "ground" insidediff --git a/README.md b/README.mdindex 66c0c64e..6f7deace 100644--- a/README.md+++ b/README.md@@ -9,9 +9,9 @@ for Nano Banana, ChatGPT, ComfyUI, Midjourney and other AI image tools.**English · [简体中文](docs/readme/README.zh-CN.md) · [日本語](docs/readme/README.ja.md) · [한국어](docs/readme/README.ko.md)-[](data/catalog.json)-[](docs/GENERATION.md)-[](docs/GENERATION.md)+[](data/catalog.json)+[](docs/GENERATION.md)+[](docs/GENERATION.md)[](data/catalog.csv)[](web)[](LICENSE-images)@@ -36,10 +36,10 @@ All images are original to this project.| Series | Poses shown | Views that passed QC | On the site ||---|---:|---:|---|-| Full body · standard (`std`, P###) | 50 of 73 | 208 / 365 | shown |-| Upper body (`upper`, U###) | 77 of 80 | 307 / 400 | shown |-| Full body · sexy (`sexy`, S###) | 80 of 80 | 341 / 400 | hidden; a switch shows it |-| Duo (`duo`, D###) | 49 of 55 | 193 / 275 | hidden; a switch shows it |+| Full body · standard (`std`, P###) | 71 of 73 | 287 / 365 | shown |+| Upper body (`upper`, U###) | 80 of 80 | 337 / 400 | shown |+| Full body · sexy (`sexy`, S###) | 80 of 80 | 323 / 400 | hidden; a switch shows it |+| Duo (`duo`, D###) | 49 of 55 | 186 / 275 | hidden; a switch shows it |Poses rated `suggestive` (the sexy and duo series, upper's Sexy and Chest emphasis) are hidden by default and kept offthe marquee, pose pages, share cards and sitemap; one switch shows them.@@ -47,10 +47,12 @@ the marquee, pose pages, share cards and sitemap; one switch shows them.- **Every view is reviewed** ([docs/qc/](docs/qc/README.md)). A view that fails (nipple marks, a face or hair,clothing, background, broken anatomy, the wrong camera angle, a figure that reads as a minor, or a drawing that is notthe named pose) is removed- from `images/` and the catalog, and `data/qc.json` keeps the reason. 23 std sets show a different pose from their- name and stay hidden until they are relabelled.+ from `images/` and the catalog, and `data/qc.json` keeps the reason.+- 43 sets were redrawn by a second generator where the original had at most two usable views or showed another pose+ (catalog `source: digi`, a slightly different mannequin with milder high and low angles; whole sets, never mixed).+ P038 and P052 still show a different pose from their name and stay hidden.- Views that were never generated stay empty rather than being filled with a bad drawing.-- **103 photo renders** for std P001–P022 and P050, reviewed like the line art: AI-generated adults, not real people.+- **89 photo renders** for std P001–P022 and P050 (P019, P020 and P022 are off: they show the old drawing), reviewed like the line art: AI-generated adults, not real people.- `data/catalog.json` / `catalog.csv`: series, rating, category and name in four languages, an English posedescription for std, the file for each view, and why a view is empty.diff --git a/docs/DECISIONS.md b/docs/DECISIONS.mdindex c1c49bfa..9d6dd10c 100644--- a/docs/DECISIONS.md+++ b/docs/DECISIONS.md@@ -4,7 +4,10 @@## 待定-### D-07 23 套 std 姿势画的和名字对不上(2026-10-28)+### D-07 23 套 std 姿势画的和名字对不上(2026-10-28;2026-10-29 大部分已解决)+2026-10-29 更新:21 套由另一个生成源按名字重画后放回(D-09),只剩 P038、P052 两套仍隐藏,新画也没通过(P038 肢体问题,P052 画的仍不是这个姿势)。+对这两套仍可二选一:按实际画面改名放回,或等下次重画。以下是原记录。+P019–P052 一段里有 23 套(P019、P020、P022、P025–P027、P029–P038、P043–P048、P052),线稿本身干净、五机位一致,但画的是另一个姿势(多是站立、行走),名字和分类都对不上。现在整套判 `structure` 不显示,`data/qc.json` 的 `sets` 里记了每套实际画的是什么。可选:1. 按实际画面重新命名、改分类(四语 + `desc_en`),复审后放回,std 回到 73 套;2. 保持隐藏,以后重新生成。@@ -12,6 +15,15 @@ P019–P052 一段里有 23 套(P019、P020、P022、P025–P027、P029–P038## 已决定+### D-09 用另一个生成源整套重画补缺(2026-10-29,人定)+comfyui-digi 交付 66 套重画(风格略有不同:蛋形头、结构线、球关节,俯视 / 仰视角度更缓)。人定:+- 风格可以混在同一个画廊里,但一套之内不混:一套要么全是旧画,要么全是新画;+- 旧画优先:新画按本仓库的质检流程通过的机位比旧画多,才整套换(最终 43 套,`source: "digi"`),否则保留旧画;+- 这 43 套的 originals 是 900×1200,照此发布;+- 按旧画渲染、和新线稿对不上的实拍(P019、P020、P022)下架;+- 分级检查给出的 6 处不一致不改,沿用分类规则。+细节见 [GENERATION.md](GENERATION.md) 和 [qc/README.md](qc/README.md) 第三轮。+### D-08 质检不通过的图从 git 历史里去掉(2026-10-29,人定)人定「可以 rewrite」。用 `git filter-repo --invert-paths` 去掉历史里所有不在当前 HEAD 的 `images/` 路径(346 个文件:std 156 个机位 + 双人 8 个儿童比例机位,各两档;实拍 9 张,两档),HEAD 的 tree 前后完全一致,然后 `--force-with-lease` 推到 origin(旧 `02d51f5` → 新 `443c368`)。diff --git a/docs/GENERATION.md b/docs/GENERATION.mdindex 2c8260be..44d4a9a8 100644--- a/docs/GENERATION.md+++ b/docs/GENERATION.md@@ -18,6 +18,18 @@- 成品 252/275,生成阶段缺 23 格(四个系列里被拒率最高,双人接触 + 正面机位最甚);D022 整套未成。明细见源包 README,已记成 `gaps: missing`。- 全系 `rating: suggestive`。+## 重画补缺 digi(2026-10-29,43 套)+另一个会话(comfyui-digi)用同一份姿势目录、同一个基准素体重画了 66 套 shisei 缺图或画错的套,交付 290 格 900×1200 PNG+(含每套四语标签和它自己的质检结果,只作参考)。素体是蛋形头、带结构线和球关节,比原画略丰满;俯视约 30°,仰视约 45° 偏左低角度。+- 暂存:`tools/import_redraws.py` 无损转成 `~/Downloads/pose-gallery-digi66/img/<series>_<ID>_<view>.webp`(源包 index、README、labels 一并拷过去,只读),+ catalog 里这些姿势加 `source: "digi"`、整套换成新画的机位。名字、分类、分级、`orig_id` 都不动(ja 为准,不用对方的译名)。+- 质检:按 [qc/README.md](qc/README.md) 第三轮重新审,对方的结果不入库。+- 选套(人定):旧画优先,新画通过的机位比旧画多才整套换,不按机位混用(五个机位必须是同一张画的同一个姿势)。66 套里 43 套换:+ P019、P020、P022、P025、P026、P027、P029、P030、P031、P032、P033、P034、P035、P036、P037、P040、P043、P044、P045、P046、P047、P048、P054、P073、S011、S017、S036、S037、S039、S052、S065、U004、U007、U017、U026、U028、U037、U038、U051、U052、U053、U072、U078。其余 23 套保留旧画(含 P038:新画两格过了,但肢体复核后为 0,和旧画一样)。+- 分辨率:这 43 套的「原图」是 900×1200,Release 的 originals 包里照此发布。+- 实拍:P019、P020、P022 的实拍是按旧画(另一个姿势)渲染的,人定下架(`photo_manual.json`),源包里仍保留。+- 发布:`build_images.py` / `pack_release.py` / `review_sheets.py` 要多给一个 `digi=~/Downloads/pose-gallery-digi66`。+## 质检四个系列的线稿和实拍都过了质检,方法和结果见 [qc/README.md](qc/README.md)。diff --git a/docs/qc/README.md b/docs/qc/README.mdindex cf359d98..c46d539b 100644--- a/docs/qc/README.md+++ b/docs/qc/README.md@@ -99,9 +99,21 @@ S068 正面 / 侧面 / 仰视(顶部横线),U032 仰视(桌沿线),D- 改成性感:P057、P071(因此不再有静态姿势页和分享卡)。其余 59 处保持按分类分级。+## 第三轮(2026-10-29):重画补缺 digi+另一个生成源重画的 66 套(290 格,docs/GENERATION.md),按本流程从头审,生成源自己的质检不算数:+- 逐格:3 个 Sonnet 审 122 张审图(`fail_digi_a?.json`),64 格不过;多数是「俯视 / 仰视其实是平视」或和正面几乎一样。+ 姿势名:2 个 Sonnet(`name_digi_a?.json`),51 yes / 12 partial / 3 no(D024、P052、S068)。`line_dups.py --source digi`:0 个重复。+- 人定只换「新画通过的机位比旧画多」的套,最终 43 套(对照页逐套给人看过)。对这 43 套已通过的机位再跑:+ 肢体计数 3 个 Sonnet(`limbs_digi_a?.json`,标 2 格,人看后加 P038 正面共下架 3 格,写在 `fail_digi_manual.json`);+ 儿童比例 1 个 Sonnet,0 格;分级 1 个 Sonnet(`rating_digi_aa.json`),6 套和分类规则不一致,人定不改。+- 合并规则:结果文件顶层写 `"source": "digi"` 的只作用于 catalog 里同一 `source` 的姿势,旧文件只作用于原画,所以两套结果都留在+ `data/qc-input/` 里,`qc_merge.py` 照样能整目录重算;新画有机位没出现在任何一张已审审图上、或缺姿势名检查时,`qc_merge.py` 拒绝合并。+ `qc.json` 新增 `sources`,记每个重画姿势的来源,`verify.py` 核对它和 catalog 一致。+- 结果:std 71/73 套、287 格;upper 80/80、337 格;sexy 80/80、323 格;duo 不变。实拍 89/112(P019、P020、P022 按旧画拍的 14 张人定下架)。+## 复审 / 新系列```bash-SRC="std=$HOME/Downloads/pose-gallery-std73 sexy=$HOME/Downloads/pose-gallery-sexy80 upper=$HOME/Downloads/pose-gallery-upper80 duo=$HOME/Downloads/pose-gallery-duo55"+SRC="std=$HOME/Downloads/pose-gallery-std73 sexy=$HOME/Downloads/pose-gallery-sexy80 upper=$HOME/Downloads/pose-gallery-upper80 duo=$HOME/Downloads/pose-gallery-duo55 digi=$HOME/Downloads/pose-gallery-digi66"python3 tools/review_sheets.py /tmp/qc/panels --mode panels ${=SRC} # zsh: ${=SRC} 才会按空格拆开;只给要审的系列ls /tmp/qc/panels/*.jpg | split -l 35 - /tmp/qc/panels/batch_# 每个批次派一个 qc-reviewer:提示词 docs/qc/review_prompt.md,输出 data/qc-input/fail_r3_<批次>.jsondiff --git a/docs/readme/README.ja.md b/docs/readme/README.ja.mdindex 9429fb8a..bdc9de40 100644--- a/docs/readme/README.ja.md+++ b/docs/readme/README.ja.md@@ -8,9 +8,9 @@[English](../../README.md) · [简体中文](README.zh-CN.md) · 日本語 · [한국어](README.ko.md)-[](../../data/catalog.json)-[](../GENERATION.md)-[](../GENERATION.md)+[](../../data/catalog.json)+[](../GENERATION.md)+[](../GENERATION.md)[](../../data/catalog.csv)[](../../web)[](../../LICENSE-images)@@ -32,16 +32,17 @@ Shisei は自作のポーズ参照ライブラリです。すべてのポーズ| シリーズ | 表示ポーズ | 品質チェック合格 | サイトでの扱い ||---|---:|---:|---|-| 全身・標準(`std`、P###) | 50 / 73 | 208 / 365 | 表示 |-| 上半身(`upper`、U###) | 77 / 80 | 307 / 400 | 表示 |-| 全身・セクシー(`sexy`、S###) | 80 / 80 | 341 / 400 | 非表示(スイッチで表示できます) |-| ふたり(`duo`、D###) | 49 / 55 | 193 / 275 | 非表示(スイッチで表示できます) |+| 全身・標準(`std`、P###) | 71 / 73 | 287 / 365 | 表示 |+| 上半身(`upper`、U###) | 80 / 80 | 337 / 400 | 表示 |+| 全身・セクシー(`sexy`、S###) | 80 / 80 | 323 / 400 | 非表示(スイッチで表示できます) |+| ふたり(`duo`、D###) | 49 / 55 | 186 / 275 | 非表示(スイッチで表示できます) |`suggestive` のポーズ(全身・セクシー、ふたり、上半身のセクシー系・胸元強調系)は初期状態で非表示で、トップのスクロール、ポーズページ、シェアカード、sitemap には入りません。スイッチひとつでまとめて表示できます。-- **すべてのアングルを品質チェック**しています([docs/qc/](../qc/README.md))。乳首の印、顔や髪、衣服、背景、崩れた構造、違うアングル、未成年に見える人物、名前と違うポーズのものは `images/` と catalog から外し、理由を `data/qc.json` に残しています。std の 23 ポーズは名前と絵が合わないため、名前を直すまで非表示です。+- **すべてのアングルを品質チェック**しています([docs/qc/](../qc/README.md))。乳首の印、顔や髪、衣服、背景、崩れた構造、違うアングル、未成年に見える人物、名前と違うポーズのものは `images/` と catalog から外し、理由を `data/qc.json` に残しています。+- 元の絵が 2 アングル以下、または名前と違うポーズだった 43 セットは、別の生成元で描き直しました(catalog の `source: digi`。少し違うマネキンで、俯瞰・あおりの角度は控えめ。セット単位で差し替え、混在はさせません)。P038 と P052 は名前と絵が合わないため非表示のままです。- 生成できなかったアングルは、崩れた絵で埋めずに空けています。-- **実写 103 枚**(std の P001–P022 と P050)。線画と同じく品質チェック済みです。AI 生成の成人で、実在の人物ではありません。+- **実写 89 枚**(std の P001–P022 と P050。P019・P020・P022 は旧い絵のポーズなので外しています)。線画と同じく品質チェック済みです。AI 生成の成人で、実在の人物ではありません。- `data/catalog.json` / `catalog.csv`:シリーズ、レーティング、4 言語の分類と名前、std の英語ポーズ説明、アングルごとのファイル、空欄の理由。## 使い方diff --git a/docs/readme/README.ko.md b/docs/readme/README.ko.mdindex 8103f92e..e8fc3228 100644--- a/docs/readme/README.ko.md+++ b/docs/readme/README.ko.md@@ -8,9 +8,9 @@[English](../../README.md) · [简体中文](README.zh-CN.md) · [日本語](README.ja.md) · 한국어-[](../../data/catalog.json)-[](../GENERATION.md)-[](../GENERATION.md)+[](../../data/catalog.json)+[](../GENERATION.md)+[](../GENERATION.md)[](../../data/catalog.csv)[](../../web)[](../../LICENSE-images)@@ -32,16 +32,17 @@ Shisei는 직접 만든 포즈 참고 라이브러리입니다. 모든 포즈는| 시리즈 | 표시 포즈 | 품질 검사 통과 앵글 | 사이트에서 ||---|---:|---:|---|-| 전신・표준 (`std`, P###) | 50 / 73 | 208 / 365 | 표시 |-| 상반신 (`upper`, U###) | 77 / 80 | 307 / 400 | 표시 |-| 전신・섹시 (`sexy`, S###) | 80 / 80 | 341 / 400 | 숨김 (스위치로 표시할 수 있음) |-| 2인 (`duo`, D###) | 49 / 55 | 193 / 275 | 숨김 (스위치로 표시할 수 있음) |+| 전신・표준 (`std`, P###) | 71 / 73 | 287 / 365 | 표시 |+| 상반신 (`upper`, U###) | 80 / 80 | 337 / 400 | 표시 |+| 전신・섹시 (`sexy`, S###) | 80 / 80 | 323 / 400 | 숨김 (스위치로 표시할 수 있음) |+| 2인 (`duo`, D###) | 49 / 55 | 186 / 275 | 숨김 (스위치로 표시할 수 있음) |`suggestive` 등급 포즈(전신・섹시, 2인, 상반신의 섹시계·가슴 강조계)는 기본으로 숨겨지며, 첫 화면 스크롤, 포즈 페이지, 공유 카드, sitemap에는 들어가지 않습니다. 스위치 하나로 모두 표시할 수 있습니다.-- **모든 앵글을 품질 검사**했습니다([docs/qc/](../qc/README.md)). 유두 표시, 얼굴이나 머리카락, 옷, 배경, 망가진 구조, 잘못된 앵글, 미성년자로 보이는 인물, 이름과 다른 포즈는 `images/`와 catalog에서 뺐고, 이유는 `data/qc.json`에 남겼습니다. std의 23개 세트는 그림이 이름과 달라 이름을 고칠 때까지 숨깁니다.+- **모든 앵글을 품질 검사**했습니다([docs/qc/](../qc/README.md)). 유두 표시, 얼굴이나 머리카락, 옷, 배경, 망가진 구조, 잘못된 앵글, 미성년자로 보이는 인물, 이름과 다른 포즈는 `images/`와 catalog에서 뺐고, 이유는 `data/qc.json`에 남겼습니다.+- 원래 그림이 2앵글 이하였거나 이름과 다른 포즈였던 43개 세트는 다른 생성원으로 다시 그렸습니다(catalog의 `source: digi`. 약간 다른 마네킹이고 위·아래 앵글이 완만합니다. 세트 단위로 교체하며 섞지 않습니다). P038과 P052는 그림이 이름과 달라 계속 숨깁니다.- 생성하지 못한 앵글은 잘못된 그림으로 채우지 않고 비워 둡니다.-- **실사 103장** (std의 P001–P022, P050). 선화와 같이 품질 검사를 거쳤습니다. AI로 생성한 성인이며 실존 인물이 아닙니다.+- **실사 89장** (std의 P001–P022, P050. P019·P020·P022는 예전 그림의 포즈라 뺐습니다). 선화와 같이 품질 검사를 거쳤습니다. AI로 생성한 성인이며 실존 인물이 아닙니다.- `data/catalog.json` / `catalog.csv`: 시리즈, 등급, 4개 언어 분류와 이름, std의 영문 포즈 설명, 앵글별 파일, 빈 앵글의 이유.## 사용 방법diff --git a/docs/readme/README.zh-CN.md b/docs/readme/README.zh-CN.mdindex 80504ded..0b8cd062 100644--- a/docs/readme/README.zh-CN.md+++ b/docs/readme/README.zh-CN.md@@ -8,9 +8,9 @@[English](../../README.md) · 简体中文 · [日本語](README.ja.md) · [한국어](README.ko.md)-[](../../data/catalog.json)-[](../GENERATION.md)-[](../GENERATION.md)+[](../../data/catalog.json)+[](../GENERATION.md)+[](../GENERATION.md)[](../../data/catalog.csv)[](../../web)[](../../LICENSE-images)@@ -32,16 +32,17 @@ Shisei 是一套自己生成的姿势参考库。每个姿势都是无脸素体| 系列 | 显示的套数 | 质检通过的机位 | 页面上 ||---|---:|---:|---|-| 全身・标准(`std`,P###) | 50 / 73 | 208 / 365 | 默认显示 |-| 上半身(`upper`,U###) | 77 / 80 | 307 / 400 | 默认显示 |-| 全身・性感(`sexy`,S###) | 80 / 80 | 341 / 400 | 默认隐藏,开关可显示 |-| 双人(`duo`,D###) | 49 / 55 | 193 / 275 | 默认隐藏,开关可显示 |+| 全身・标准(`std`,P###) | 71 / 73 | 287 / 365 | 默认显示 |+| 上半身(`upper`,U###) | 80 / 80 | 337 / 400 | 默认显示 |+| 全身・性感(`sexy`,S###) | 80 / 80 | 323 / 400 | 默认隐藏,开关可显示 |+| 双人(`duo`,D###) | 49 / 55 | 186 / 275 | 默认隐藏,开关可显示 |标为 `suggestive` 的姿势(全身・性感、双人,上半身的性感系、胸线强调系)默认隐藏,不进首屏滚动方块、姿势页、分享卡和 sitemap,一个开关就能全部显示。-- **每个机位都过了质检**([docs/qc/](../qc/README.md))。有乳头标记、五官或头发、衣物、背景、结构畸形、机位不对、人物看起来像未成年,或者画的不是这个姿势的,都从 `images/` 和 catalog 里拿掉了,原因记在 `data/qc.json`。std 有 23 套画的姿势和名字对不上,重新命名之前先不显示。+- **每个机位都过了质检**([docs/qc/](../qc/README.md))。有乳头标记、五官或头发、衣物、背景、结构畸形、机位不对、人物看起来像未成年,或者画的不是这个姿势的,都从 `images/` 和 catalog 里拿掉了,原因记在 `data/qc.json`。+- 原画只有两个以下可用机位、或画的不是这个姿势的 43 套,用另一个生成源整套重画(catalog 里 `source: digi`;素体略有不同,俯视、仰视角度更缓;整套替换,不混用)。P038、P052 仍然和名字对不上,继续不显示。- 没生成出来的机位宁可留空,也不放一张画错的。-- **103 张实拍**,覆盖 std 的 P001–P022 和 P050,和线稿一样过了质检,都是 AI 生成的成年人物,不对应任何真人。+- **89 张实拍**,覆盖 std 的 P001–P022 和 P050(P019、P020、P022 拍的是旧画的姿势,已拿掉),和线稿一样过了质检,都是 AI 生成的成年人物,不对应任何真人。- `data/catalog.json` 和 `catalog.csv` 里有系列、分级、四语分类和名称、std 的英文姿势描述、每个机位的文件,以及机位为空的原因。## 怎么用diff --git a/web/changelog.yaml b/web/changelog.yamlindex 8f3d8467..3963770c 100644--- a/web/changelog.yaml+++ b/web/changelog.yaml@@ -4,6 +4,22 @@# tag: new | improve | announce | removed. link: an in-site path without language prefix (/, /changelog/, /pose/P001/),# optionally with ?query and #hash; the build prefixes the reader's language. Newest first.entries:+ - date: 2026-10-29+ tag: new+ title: 43 pose sets redrawn, 25 back on show+ desc: Sets with at most two usable angles or the wrong pose were redrawn whole and checked again. Standard now shows 71 of 73 sets, upper body all 80. The new mannequin looks a little different.+ link: /+ cta: Open the gallery+ title_zh: 重画 43 套姿势,25 套重新上架+ desc_zh: 原来只有两个以下可用机位、或画错姿势的套,整套重画并重新质检。标准系现在显示 73 套里的 71 套,上半身 80 套全部显示。新画的素体略有不同。+ cta_zh: 打开画廊+ title_ja: 43 セットを描き直し、25 セットが復活+ desc_ja: 使えるアングルが 2 つ以下、またはポーズが違っていたセットをセットごと描き直し、改めて品質チェックしました。標準は 73 セット中 71、上半身は 80 セットすべてを表示します。新しい絵はマネキンが少し違います。+ cta_ja: ギャラリーを開く+ title_ko: 43개 세트를 다시 그려 25개 세트 복귀+ desc_ko: 쓸 수 있는 앵글이 2개 이하이거나 포즈가 틀렸던 세트를 세트 단위로 다시 그리고 품질 검사를 다시 했습니다. 표준은 73개 중 71개, 상반신은 80개 모두 표시합니다. 새 그림은 마네킹이 조금 다릅니다.+ cta_ko: 갤러리 열기+- date: 2026-10-29tag: improvetitle: More poses per screen on phones