Skip to content

💃 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

feat(tools): briefs for several categories at once, with a skip list; record go-ahead for the new libraries

wquguru2c8ce11fbeff
diff --git a/docs/DECISIONS.md b/docs/DECISIONS.md
index da1aef13..5f58d220 100644
--- a/docs/DECISIONS.md
+++ b/docs/DECISIONS.md
@@ -4,6 +4,15 @@
## 待定
+### 2026-11-01 开工:场景、道具、家具、光线、动物(人已同意)
+人在 2026-11-01 同意五个新库全部开工(“把所有的都”),以下各条的推荐项按此生效,条目保留作记录:
+- D-12 场景、D-13 道具、D-16 家具:按 D-17 的范围、各自推荐的视角和 ID 字母开工。道具 T009 琵琶、T015 折扇试验三次未过,先不进正式轮。
+- D-15 光线:人台 + 灰球(全身四类)和石膏胸像(PO),三个平视机位,ID `L`。
+- D-14 动物:画风定为 A 线稿(与姿势库同一画法),四足动物侧面主图、鸟和龙正面主图,第二张平面视角参考主图,斜侧 / 俯视 / 仰视参考两张;
+ 主图必须通过,其余视角尽力,画不对的留 `gaps: missing`,不拿低质量图凑。完整动作卡还没写,写完才开工。
+- “带不带参考图”对照没跑,件数仍按 D-17 的判断;正式轮的通过率和用途反馈再决定加减。
+- 新图入库仍要过 QC(AGENTS §2):各库的审图提示词和收图工具还没写,图先生成在 `~/Downloads`,工具写好、审过再入库。
+
### D-17 场景、道具、家具三库:生成前重排、缩量、重新编号(2026-10-31,人已同意方向,件数暂定)
依据 2026 年生图、生视频的做法调研:参考图的价值在“文字说不清”和“多张图要对得上”;通用物件、通用地点写文字就能画好;
一次塞太多参考图会被忽略或照抄,所以组合按每次最多 3 张参考(姿势 + 服装 + 场景或光其一)设计。人同意按此重排:
diff --git a/tools/series_brief.py b/tools/series_brief.py
index 0e906215..dc30f370 100644
--- a/tools/series_brief.py
+++ b/tools/series_brief.py
@@ -4,7 +4,8 @@
The counterpart of costume_brief.py. Either one category, or --probe: the first variant of every form marked
`probe`, across all categories, to test each category before the full run (the costume-probe01 step).
-Usage: python3 -I tools/series_brief.py scene IN ~/Downloads/scene-IN-run01 [--only E001,E002] [--gate 10]
+Usage: python3 -I tools/series_brief.py scene IN ~/Downloads/scene-IN-run01 [--only E001,E002] [--skip E003] [--gate 10]
+ python3 -I tools/series_brief.py scene IN,UR ~/Downloads/scene-IN-UR-run01 (several categories, one paste)
python3 -I tools/series_brief.py prop --probe ~/Downloads/prop-probe01
Writes OUT/brief.md and OUT/manifest.json; items that already have every view are skipped. The animal library only
has its --probe brief, which draws every probe item in two styles to decide D-14 (docs/animal/PLAN.md).
@@ -223,6 +224,7 @@ def main() -> None:
ap.add_argument("cat", nargs="?", default="")
ap.add_argument("out")
ap.add_argument("--only", default="")
+ ap.add_argument("--skip", default="") # items held back, e.g. still failing their probe
ap.add_argument("--probe", action="store_true")
ap.add_argument("--gate", type=int, default=10)
a = ap.parse_args()
@@ -234,8 +236,13 @@ def main() -> None:
views = cfg["views"]
cats = json.loads((ROOT / cfg["src"] / "_categories.json").read_text())
only = set(filter(None, a.only.split(",")))
+ skip = set(filter(None, a.skip.split(",")))
+ want = set(filter(None, a.cat.split(",")))
+ if want - set(cats):
+ raise SystemExit(f"unknown category: {', '.join(sorted(want - set(cats)))}")
items = [i for i in json.loads((ROOT / cfg["out"]).read_text())["items"]
- if (i.get("probe") if a.probe else i["cat"] == a.cat) and (not only or i["id"] in only)
+ if (i.get("probe") if a.probe else i["cat"] in want) and (not only or i["id"] in only)
+ and i["id"] not in skip
and not all(i["views"].get(v) for v in views)]
if not items:
raise SystemExit("nothing to brief")
@@ -246,12 +253,14 @@ def main() -> None:
need = {i["id"]: [v for v in order(i) if not i["views"].get(v)] for i in items} # drawing order: references first
n_img = sum(len(v) for v in need.values()) + (len(items) if a.lib == "animal" else 0) # + one B-style front each
cat_list = sorted({i["cat"] for i in items}, key=lambda c: cats[c]["order"])
- title = "试验(每个大类挑几件,验证能不能稳定生成)" if a.probe else f"{cats[a.cat]['name']['zh']}({a.cat})"
+ title = ("试验(每个大类挑几件,验证能不能稳定生成)" if a.probe
+ else "、".join(f"{cats[c]['name']['zh']}({c})" for c in cat_list))
unit = {"scene": "个场景", "light": "个光线预设", "animal": "个动作"}.get(a.lib, "件")
view_zh = "、".join(VIEW_ZH[v] for v in views)
gate = ([] if a.probe else [
- f"**自检闸**:先做前 {a.gate} {unit}的第一张({VIEW_ZH[views[0]]})。通过少于 {round(a.gate * 0.7)} {unit}就停下,"
- "把问题写进 notes.md 交付,不要继续;达到了再做完其余。", ""])
+ f"**自检闸**:{'每个大类' if len(cat_list) > 1 else ''}先做前 {a.gate} {unit}的第一张({VIEW_ZH[views[0]]})。"
+ f"通过少于 {round(a.gate * 0.7)} {unit}就停下这个大类,把问题写进 notes.md,"
+ + ("接着做下一个大类" if len(cat_list) > 1 else "不要继续") + ";达到了再做完其余。", ""])
L = [f"# shisei {cfg['label']} · {title} · {run}", "",
f"共 {len(items)} {unit}、{n_img} 张(视角:{view_zh})。规格见下,每{unit}的字段已经给好,照模板拼即可。"