webkubor/charmoji-sticker-skill
charmoji
Verified微信表情开放平台静态表情投稿全流程 SOP —— 一张 IP 正面照进,整套可提交素材出。覆盖 IP 命名与简介文案、9 情绪选题、museav 中台批量出图、240×240 表情图 / 750×400 详情页横幅 / 240×240 封面 / 50×50 聊天面板图标切图、透明背景与白描边机检、含义词字数校验、赞赏三件套、付费表情(10 微信豆)、审核红线与高频拒因;上架后还能一条命令出 1080×1920 竖版推广视频(合成走 reel-kit)。触发词:微信表情、表情包、表情专辑、表情投稿、提交表情、表情形象、微信表情开放平台、sticker、表情审核、含义词、赞赏引导图、付费表情、微信豆、表情封面、聊天面板图标、表情推广视频、表情包宣传片、表情串成视频。
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 19.3 KB |
| examples/album.yml | config | 1.1 KB |
| examples/submit.md | markdown | 2.1 KB |
| references/animated.md | markdown | 8.8 KB |
| references/audit.md | markdown | 8.1 KB |
| references/ip-design.md | markdown | 4.4 KB |
| references/pitfalls.md | markdown | 15.1 KB |
| references/platform.md | markdown | 6.1 KB |
| references/specs.md | markdown | 9.7 KB |
| scripts/check_assets.pyexecutable | script | 12.9 KB |
| scripts/fit_assets.pyexecutable | script | 8.2 KB |
| scripts/gen_album.shexecutable | script | 7.5 KB |
| scripts/install_skill.pyexecutable | script | 9.0 KB |
| scripts/ip.py | script | 34.8 KB |
| scripts/lint.shexecutable | script | 4.5 KB |
| scripts/make_gif.py | script | 6.8 KB |
| scripts/make_promo.pyexecutable | script | 6.7 KB |
| scripts/make_submit.pyexecutable | script | 5.6 KB |
| scripts/new_album.shexecutable | script | 10.7 KB |
| scripts/page_tpl.py | script | 10.9 KB |
| scripts/restyle.py | script | 8.3 KB |
| templates/album.yml | config | 1.5 KB |
scripts/restyle.py
View on GitHub#!/usr/bin/env python3
"""照片型表情图 → 透明贴纸(抠图 + 重绘文字)。
restyle.py <源目录> <输出目录> --texts texts.tsv --start 9
texts.tsv 每行三列(Tab 分隔):源文件名、画面文字、含义词
B-01.png<TAB>等饭饭~<TAB>等饭饭
B-02.png<TAB>困死咯~<TAB>困死咯
输出 `<起始序号+i>-<含义词>.png`,240×240 透明 PNG。
## 为什么要重绘文字而不是保留原文字
照片型表情的文字通常压在照片背景上(毯子、桌面、墙)。抠图必然把背景连文字
一起去掉 —— 实测 birefnet 抠完只剩猫,文字和粉毯都没了。所以文字只能重绘。
重绘反而是好事:原素材往往每张字体、字号、描边都不一致(有白字黑边、有黑字、
有的字特别小),重绘之后整套统一。
## 为什么要抠透明
官方对表情图**不强制**透明背景(只强制封面与图标)。但不透明的白底照片发到微信
就是一个白方块,深色模式下边缘生硬 —— 用户会描述成「有白色毛边」「不像贴纸」。
注意这个问题**查文件参数查不出来**(尺寸、四边像素全正常),必须贴到深色底上才现形。
"""
import argparse
import os
import shutil
import subprocess
import sys
from PIL import Image, ImageDraw, ImageFilter, ImageFont
FONT = "/System/Library/Fonts/Hiragino Sans GB.ttc"
FONT_INDEX = 2 # W6 粗体 —— 表情包文字要够粗才压得住浅色背景
SIDE = 240
TEXT_BAND = 0.22 # 文字占底部比例
FILL = (255, 255, 255, 255)
STROKE = (74, 58, 52, 255)
def load_font(size):
for path, idx in ((FONT, FONT_INDEX), (FONT, 0), ("/System/Library/Fonts/STHeiti Medium.ttc", 0)):
try:
return ImageFont.truetype(path, size, index=idx)
except Exception:
continue
return ImageFont.load_default()
def text_mask(path, top_ratio=0.55, dilate=17):
"""定位画面里的描边文字。特征是「白笔画」与「深描边」在小邻域内共存。
只在下半部找 —— 猫眼睛也是高对比黑白,整幅搜会把眼睛当成文字擦掉。"""
from PIL import ImageChops
im = Image.open(path).convert("L")
w, h = im.size
white = im.point(lambda v: 255 if v > 236 else 0)
dark = im.point(lambda v: 255 if v < 100 else 0)
both = ImageChops.multiply(white.filter(ImageFilter.MaxFilter(9)),
dark.filter(ImageFilter.MaxFilter(9)))
m = both.filter(ImageFilter.MaxFilter(dilate))
px = m.load()
for y in range(int(h * top_ratio)):
for x in range(w):
px[x, y] = 0
return m
def erase_text(src, dst, work):
"""擦掉原画面文字。顺序很关键:必须先擦字再抠图。
反过来的话,压在主体身上的文字会随主体一起保留,跟重绘的新文字叠在一起。
museav 的自动定位只认角标式半透明水印,这种大号描边文字要自己给 mask。"""
if not shutil.which("museav"):
return False
mask = os.path.join(work, os.path.basename(src).replace(".png", "-mask.png"))
m = text_mask(src)
if not m.getbbox():
shutil.copy2(src, dst) # 没检测到文字,原图直接过
return True
m.save(mask)
r = subprocess.run(["museav", "remove-watermark", src, "--mask", mask, "--out", dst, "--overwrite"],
capture_output=True, text=True)
if r.returncode != 0 or not os.path.exists(dst):
print(f"⚠️ 擦字失败 {os.path.basename(src)}:{(r.stdout + r.stderr).strip()[:140]}", file=sys.stderr)
shutil.copy2(src, dst)
return True
return True
def cutout(src, dst):
"""抠图。museav remove-bg 默认 birefnet —— 毛发边缘明显好于 isnet/u2net。"""
if not shutil.which("museav"):
print("❌ 需要 museav CLI 做抠图(或自己抠好后用 --skip-cutout)", file=sys.stderr)
sys.exit(1)
r = subprocess.run(["museav", "remove-bg", src, "--model", "birefnet", "--out", dst, "--overwrite"],
capture_output=True, text=True)
if r.returncode != 0 or not os.path.exists(dst):
print(f"⚠️ 抠图失败 {os.path.basename(src)}:{(r.stdout + r.stderr).strip()[:160]}", file=sys.stderr)
return False
return True
def compose(nobg, text, side=SIDE, pad=0.92):
im = Image.open(nobg).convert("RGBA")
box = im.getchannel("A").point(lambda v: 255 if v > 16 else 0).getbbox()
if box:
im = im.crop(box)
canvas = Image.new("RGBA", (side, side), (0, 0, 0, 0))
band = int(side * TEXT_BAND) if text else 0
art_h = side - band
scale = min(int(side * pad) / im.width, int(art_h * 0.98) / im.height)
im = im.resize((max(1, round(im.width * scale)), max(1, round(im.height * scale))), Image.LANCZOS)
canvas.paste(im, ((side - im.width) // 2, max(0, art_h - im.height)), im)
if text:
# 按「实际渲染宽度」往下收字号,而不是按字数估算 ——
# 估算会让「开心到飞起!」这种长句超出画面被裁掉。
d = ImageDraw.Draw(canvas)
max_w = side * 0.94
size = int(band * 0.92)
while size > 10:
f = load_font(size)
stroke = max(2, size // 9)
if d.textlength(text, font=f) + stroke * 2 <= max_w:
break
size -= 1
f = load_font(size)
stroke = max(2, size // 9)
tw = d.textlength(text, font=f)
d.text(((side - tw) / 2, side - band - 2), text, font=f, fill=FILL,
stroke_width=stroke, stroke_fill=STROKE)
return canvas
def save(im, path, max_kb=500):
im.save(path, "PNG", optimize=True)
for colors in (256, 192, 128):
if os.path.getsize(path) / 1024 <= max_kb:
break
im.quantize(colors=colors, method=Image.FASTOCTREE).save(path, "PNG", optimize=True)
def main():
ap = argparse.ArgumentParser()
ap.add_argument("src", help="原稿目录")
ap.add_argument("out", help="输出目录")
ap.add_argument("--texts", required=True, help="TSV:源文件名 / 画面文字 / 含义词")
ap.add_argument("--start", type=int, default=1, help="输出起始编号,默认 1")
ap.add_argument("--skip-cutout", action="store_true", help="源图已是透明 PNG,跳过擦字与抠图")
ap.add_argument("--keep-text", action="store_true", help="不擦原画面文字(原文字压在主体上时会与新文字重叠)")
ap.add_argument("--workdir", help="抠图中间件目录,默认放输出目录的 .cut/")
args = ap.parse_args()
rows = []
for line in open(args.texts, encoding="utf-8"):
line = line.rstrip("\n")
if not line.strip() or line.startswith("#"):
continue
parts = line.split("\t")
if len(parts) < 3:
print(f"⚠️ 跳过格式不对的行:{line[:60]}", file=sys.stderr)
continue
rows.append((parts[0].strip(), parts[1].strip(), parts[2].strip()))
os.makedirs(args.out, exist_ok=True)
work = args.workdir or os.path.join(args.out, ".cut")
os.makedirs(work, exist_ok=True)
print(f"共 {len(rows)} 张,起始编号 {args.start:02d}")
ok = 0
for i, (fname, text, word) in enumerate(rows):
src = os.path.join(args.src, fname)
if not os.path.exists(src):
print(f" ✗ 源图不存在:{fname}", file=sys.stderr)
continue
if args.skip_cutout:
nobg = src
else:
clean = os.path.join(work, "clean-" + fname)
if not (os.path.exists(clean) and os.path.getsize(clean) > 1000):
if not args.keep_text:
erase_text(src, clean, work)
else:
shutil.copy2(src, clean)
nobg = os.path.join(work, fname)
if not (os.path.exists(nobg) and os.path.getsize(nobg) > 1000):
if not cutout(clean, nobg):
continue
dst = os.path.join(args.out, f"{args.start + i:02d}-{word}.png")
save(compose(nobg, text), dst)
print(f" ✓ {os.path.basename(dst)} 「{text}」 {os.path.getsize(dst)/1024:.0f}KB")
ok += 1
print(f"\n完成 {ok}/{len(rows)} 张 → {args.out}")
print("抠图中间件留在 " + work + "(重跑会复用,删掉即重新抠)")
return 0 if ok == len(rows) else 1
if __name__ == "__main__":
sys.exit(main())
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 196k | Updated | Claude CodeCodex | Verified | Install |