Ethanfel/ComfyUI-Ethanfel-Prompt-Builder
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Extract SDXL tag policy

Browse Source
This commit is contained in:
Ethanfel
2026-06-27 11:48:54 +02:00
parent 8fc3abc504
commit 2605fae3eb
5 changed files with 325 additions and 196 deletions
Download Patch File Download Diff File Expand all files Collapse all files
sdxl_formatter.py
+18 -189
View File
@@ -1,17 +1,16 @@
from __future__ import annotations
import re
from typing import Any
try:
from . import formatter_input as input_policy
from . import route_metadata as route_metadata_policy
from . import sdxl_tag_policy
from . import sdxl_tag_routes
from . import sdxl_presets as sdxl_policy
from .prompt_hygiene import sanitize_negative_text, sanitize_tag_prompt
except ImportError: # Allows local smoke tests with `python -c`.
import formatter_input as input_policy
import route_metadata as route_metadata_policy
import sdxl_tag_policy
import sdxl_tag_routes
import sdxl_presets as sdxl_policy
from prompt_hygiene import sanitize_negative_text, sanitize_tag_prompt
@@ -69,238 +68,68 @@ def _strip_prompt_field_labels(text: str) -> str:
return input_policy.strip_prompt_field_labels(text, field_labels=PROMPT_FIELD_LABELS)
def _prompt_field(text: str, label: str) -> str:
return input_policy.prompt_field(text, label, field_labels=PROMPT_FIELD_LABELS)
def _row_value(row: dict[str, Any], key: str, labels: tuple[str, ...] = ()) -> str:
return input_policy.row_value(row, key, labels, field_labels=PROMPT_FIELD_LABELS)
def _split_tag_text(text: Any) -> list[str]:
text = _clean(text)
if not text:
return []
text = re.sub(r"\bWoman [A-Z]'s\b", "woman's", text)
text = re.sub(r"\bMan [A-Z]'s\b", "man's", text)
text = re.sub(r"\bWoman [A-Z]\b", "woman", text)
text = re.sub(r"\bMan [A-Z]\b", "man", text)
text = re.sub(
r"\b(?:Clothing state|Visual clothing state|visible remaining styling|teaser outfit detail|softcore visual reference|Sexual scene|Role graph):\s*",
"",
text,
flags=re.IGNORECASE,
)
text = re.sub(r"\b(?:and|with)\b", ",", text, flags=re.IGNORECASE)
parts = re.split(r"\s*[,;]\s*", text)
return [_clean(part).strip(" .") for part in parts if _clean(part).strip(" .")]
return sdxl_tag_policy.split_tag_text(text)
def _tag_key(tag: str) -> str:
text = _clean(tag).lower()
text = re.sub(r"^\((.*?):[0-9.]+\)$", r"\1", text)
text = text.strip("() ")
return text
return sdxl_tag_policy.tag_key(tag)
def _add(tags: list[str], seen: set[str], value: Any) -> None:
for tag in _split_tag_text(value):
key = _tag_key(tag)
if key and key not in seen:
tags.append(tag)
seen.add(key)
sdxl_tag_policy.add(tags, seen, value)
def _add_one(tags: list[str], seen: set[str], tag: str) -> None:
tag = _clean(tag).strip(" ,")
key = _tag_key(tag)
if tag and key and key not in seen:
tags.append(tag)
seen.add(key)
sdxl_tag_policy.add_one(tags, seen, tag)
def _metadata_family_tags(row: dict[str, Any]) -> list[str]:
tags: list[str] = []
action_family = route_metadata_policy.row_action_family(row)
tags.extend(SDXL_ACTION_FAMILY_TAGS.get(action_family, ()))
position_family = route_metadata_policy.row_position_family(row)
tags.extend(SDXL_POSITION_FAMILY_TAGS.get(position_family, ()))
for key in route_metadata_policy.row_position_keys(row, include_unknown=True):
key_text = _clean(key)
if key_text:
tags.append(key_text.replace("_", " "))
return tags
return sdxl_tag_policy.metadata_family_tags(row)
def _formatter_hint_tags(*rows: dict[str, Any]) -> list[str]:
tags: list[str] = []
for row in rows:
if not isinstance(row, dict):
continue
for hint in route_metadata_policy.row_formatter_hints(row, "sdxl"):
hint = _clean(hint).strip(" ,.")
if hint and hint not in tags:
tags.append(hint)
return tags
return sdxl_tag_policy.formatter_hint_tags(*rows)
def _combine_tags(*parts: Any) -> str:
tags: list[str] = []
seen: set[str] = set()
for part in parts:
_add(tags, seen, part)
return ", ".join(tags)
return sdxl_tag_policy.combine_tags(*parts)
def _combine_negative(*parts: Any) -> str:
return _combine_tags(*(part for part in parts if _clean(part)))
return sdxl_tag_policy.combine_negative(*parts)
def _count_tag(women_count: int = 0, men_count: int = 0) -> list[str]:
tags = []
if women_count > 0:
tags.append(f"{women_count}woman" if women_count == 1 else f"{women_count}women")
if men_count > 0:
tags.append(f"{men_count}man" if men_count == 1 else f"{men_count}men")
return tags
return sdxl_tag_policy.count_tag(women_count, men_count)
def _infer_counts(row: dict[str, Any]) -> tuple[int, int]:
try:
women = int(row.get("women_count") or 0)
men = int(row.get("men_count") or 0)
except (TypeError, ValueError):
women = men = 0
if women or men:
return women, men
subject = _clean(row.get("subject_type") or row.get("primary_subject")).lower()
phrase = _clean(row.get("subject_phrase")).lower()
text = f"{subject} {phrase}"
if "two women" in text:
return 2, 0
if "two men" in text:
return 0, 2
if "woman and" in text or "woman a" in text and "man a" in text:
return 1, 1
if "group" in text:
return 2, 2
if "man" in text and "woman" not in text:
return 0, 1
return 1, 0
return sdxl_tag_policy.infer_counts(row)
def _character_tags_from_descriptor(descriptor: Any) -> list[str]:
text = _clean(descriptor)
text = re.sub(r"\bWoman [A-Z]\s*/\s*primary creator:\s*", "", text)
text = re.sub(r"\b(?:Woman|Man) [A-Z]:\s*", "", text)
text = re.sub(r"\balongside\b", ",", text, flags=re.IGNORECASE)
parts = _split_tag_text(text)
cleaned = []
for part in parts:
part = re.sub(r"\bfigure\b", "build", part, flags=re.IGNORECASE)
part = part.replace("adult adult", "adult")
cleaned.append(part)
return cleaned
return sdxl_tag_policy.character_tags_from_descriptor(descriptor)
def _normal_character_tags(row: dict[str, Any]) -> list[str]:
descriptor = (
_clean(row.get("cast_descriptor_text"))
or _prompt_field(row.get("prompt", ""), "Characters")
or _prompt_field(row.get("prompt", ""), "Cast descriptors")
)
if descriptor:
return _character_tags_from_descriptor(descriptor)
parts = [
_clean(row.get("age") or row.get("age_band")),
_clean(row.get("subject_phrase") or row.get("subject_type") or row.get("primary_subject")),
_clean(row.get("body_phrase") or row.get("body") or row.get("body_type")),
_clean(row.get("skin")),
_clean(row.get("hair")),
_clean(row.get("eyes")),
]
return [part for part in parts if part and part not in ("woman", "man", "single_any")]
return sdxl_tag_policy.normal_character_tags(row)
def _camera_tags_from_config(config: Any) -> list[str]:
if not isinstance(config, dict):
return []
if _clean(config.get("camera_detail")) == "off" or _clean(config.get("camera_mode")) == "disabled":
return []
custom = _clean(config.get("custom_camera_prompt"))
tags = _split_tag_text(custom)
direction = _clean(config.get("orbit_direction"))
elevation = _clean(config.get("orbit_elevation_label"))
distance = _clean(config.get("orbit_distance_label"))
for value in (direction, elevation, distance):
if value and value != "auto":
tags.extend(_split_tag_text(value))
for key in ("angle", "shot_size", "distance", "lens", "orientation", "subject_focus"):
value = _clean(config.get(key)).replace("_", " ")
if value and value != "auto":
tags.append(value)
return tags
return sdxl_tag_policy.camera_tags_from_config(config)
def _camera_tags(row: dict[str, Any], directive: Any = "", config: Any = None) -> list[str]:
tags = _split_tag_text(directive)
tags.extend(_camera_tags_from_config(config if config is not None else row.get("camera_config")))
camera_directive = _clean(row.get("camera_directive"))
if camera_directive:
tags.extend(_split_tag_text(camera_directive))
out = []
for tag in tags:
tag = tag.replace("0-degree front view", "(front facing:1.15)")
tag = tag.replace("front view", "(front facing:1.15)")
tag = tag.replace("right side view", "side view")
tag = tag.replace("left side view", "side view")
out.append(tag)
return out
return sdxl_tag_policy.camera_tags(row, directive, config)
def _explicit_tags(text: str, nude_weight: float) -> list[str]:
lower = text.lower()
tags: list[str] = []
if any(token in lower for token in ("fully nude", "fully exposed", "naked", "bare skin unobstructed", "explicit_nude")):
tags.append(f"(naked:{nude_weight:.2f})")
if any(token in lower for token in ("nipples", "breasts exposed", "bare breasts", "nipple")):
tags.append("nipples")
if any(token in lower for token in ("pussy", "vulva", "genitals")):
tags.append("pussy")
if any(token in lower for token in ("penis", "cock")):
tags.append("penis")
if "penetration" in lower or "thrust" in lower:
tags.append("penetration")
if "vaginal" in lower:
tags.append("pussy")
if "oral" in lower or "mouth" in lower:
tags.append("oral sex")
if "anal" in lower:
tags.append("anal sex")
if any(token in lower for token in ("semen", "ejaculates", "cum ")):
tags.append("semen")
return tags
return sdxl_tag_policy.explicit_tags(text, nude_weight)
def _sdxl_tag_route_dependencies() -> sdxl_tag_routes.SDXLTagRouteDependencies:
return sdxl_tag_routes.SDXLTagRouteDependencies(
clean=_clean,
row_value=_row_value,
tag_key=_tag_key,
add=_add,
add_one=_add_one,
count_tag=_count_tag,
infer_counts=_infer_counts,
normal_character_tags=_normal_character_tags,
character_tags_from_descriptor=_character_tags_from_descriptor,
metadata_family_tags=_metadata_family_tags,
formatter_hint_tags=_formatter_hint_tags,
camera_tags=_camera_tags,
explicit_tags=_explicit_tags,
)
return sdxl_tag_policy.tag_route_dependencies()
def _row_core_tags(row: dict[str, Any], nude_weight: float) -> list[str]: