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

Generalize scene camera profiles

Browse Source
This commit is contained in:
Ethanfel
2026-06-27 13:13:31 +02:00
parent 616d1132ff
commit 63e8489fb2
4 changed files with 252 additions and 83 deletions
Download Patch File Download Diff File Expand all files Collapse all files
scene_camera_adapters.py
+209 -70
View File
@@ -25,26 +25,120 @@ CAMERA_DISTANCES = (
)
def is_coworking_scene(scene_text: Any) -> bool:
text = str(scene_text or "").lower()
return any(
term in text
for term in (
SCENE_CAMERA_PROFILES: tuple[dict[str, Any], ...] = (
{
"key": "business_cafe",
"family": "coworking",
"terms": ("business cafe", "work cafe", "cafe"),
"layout_label": "Business cafe camera layout",
"place": "business cafe coworking counter",
"foreground": "counter edge, laptop corner, and small plant",
"midground": "bar stools, warm desk lamps, and coffee-counter work spots",
"background": "plants, mirror strip, menu wall, and repeated cafe work tables",
"detail_label": "cafe details",
"composition": {
"woman": "business-cafe selfie frame with the woman near a counter edge and warm work-table depth behind her",
"man": "business-cafe portrait frame with the man near a counter edge and warm work-table depth behind him",
"default": "business-cafe frame with the subjects near a counter edge and warm work-table depth behind them",
},
},
{
"key": "office_after_hours",
"family": "coworking",
"terms": ("corporate office", "office after hours", "copier", "office lounge"),
"layout_label": "Office camera layout",
"place": "empty after-hours office",
"foreground": "copier alcove edge, chair backs, and nearest desk corner",
"midground": "repeating desks, glass partition seams, and muted monitor glow",
"background": "rows of empty workstations, city-light windows, and quiet office depth",
"detail_label": "office details",
"composition": {
"woman": "after-hours office frame with the woman near a desk edge and glass-partition depth behind her",
"man": "after-hours office frame with the man near a desk edge and glass-partition depth behind him",
"default": "after-hours office frame with the subjects near a desk edge and glass-partition depth behind them",
},
},
{
"key": "coworking_lounge",
"family": "coworking",
"terms": (
"coworking",
"cowork",
"office lounge",
"business cafe",
"work cafe",
"shared office",
"corporate office",
"office after hours",
"laptops",
"warm desks",
"repeating desks",
"glass partitions",
"copier alcove",
)
)
),
"layout_label": "Coworking camera layout",
"place": "coworking lounge",
"foreground": "near desk edge, laptop corner, and chair back",
"midground": "warm work desks, laptop tables, and glass partition seams",
"background": "tall windows, repeated desk rows, plants, and soft shared-office depth",
"detail_label": "coworking details",
"composition": {
"woman": "coworking lounge selfie frame with the woman near a desk edge and tall-window depth behind her",
"man": "coworking lounge portrait frame with the man near a desk edge and tall-window depth behind him",
"default": "coworking lounge frame with the subjects near a desk edge and tall-window depth behind them",
},
},
{
"key": "classical_library",
"family": "library",
"terms": (
"classical library",
"library stacks",
"large library",
"grand library",
"reading room",
"bookshelves",
"book shelves",
"book stacks",
"rare-books",
"rare books",
"rolling ladders",
),
"layout_label": "Library camera layout",
"place": "classical library",
"foreground": "near bookshelf edge, reading-table corner, and brass lamp",
"midground": "towering bookshelves, rolling ladders, carved columns, and marble floor lines",
"background": "arched windows, repeated book aisles, warm brass lamps, and deep quiet library depth",
"detail_label": "library details",
"composition": {
"woman": "classical library frame with the woman near a bookshelf edge and long shelf depth behind her",
"man": "classical library frame with the man near a bookshelf edge and long shelf depth behind him",
"default": "classical library frame with the subjects near a bookshelf edge and long shelf depth behind them",
},
},
)
MISMATCHED_COMPOSITION_TERMS = (
"outfit-check",
"outfit check",
"mirror view",
"mirror pose",
"bag",
"shoes",
"footwear",
)
def scene_camera_profile(scene_text: Any) -> dict[str, Any]:
text = str(scene_text or "").lower()
if not text:
return {}
for profile in SCENE_CAMERA_PROFILES:
if any(term in text for term in profile["terms"]):
return dict(profile)
return {}
def is_coworking_scene(scene_text: Any) -> bool:
return scene_camera_profile(scene_text).get("family") == "coworking"
def is_scene_camera_aware(scene_text: Any) -> bool:
return bool(scene_camera_profile(scene_text))
def _compact_label(value: Any, compact_labels: Mapping[str, str] | None = None) -> str:
@@ -97,30 +191,10 @@ def camera_distance_from_text(text: Any) -> str:
def coworking_location_profile(scene_text: Any) -> dict[str, str]:
text = str(scene_text or "").lower()
if "business cafe" in text or "work cafe" in text or "cafe" in text:
return {
"layout_label": "Business cafe camera layout",
"place": "business cafe coworking counter",
"foreground": "counter edge, laptop corner, and small plant",
"midground": "bar stools, warm desk lamps, and coffee-counter work spots",
"background": "plants, mirror strip, menu wall, and repeated cafe work tables",
}
if "corporate office" in text or "office after hours" in text or "copier" in text:
return {
"layout_label": "Office camera layout",
"place": "empty after-hours office",
"foreground": "copier alcove edge, chair backs, and nearest desk corner",
"midground": "repeating desks, glass partition seams, and muted monitor glow",
"background": "rows of empty workstations, city-light windows, and quiet office depth",
}
return {
"layout_label": "Coworking camera layout",
"place": "coworking lounge",
"foreground": "near desk edge, laptop corner, and chair back",
"midground": "warm work desks, laptop tables, and glass partition seams",
"background": "tall windows, repeated desk rows, plants, and soft shared-office depth",
}
profile = scene_camera_profile(scene_text)
if profile.get("family") == "coworking":
return profile
return scene_camera_profile("coworking lounge")
def coworking_subject_terms(subject_kind: str, pov_labels: list[str] | None = None) -> tuple[str, str]:
@@ -135,7 +209,7 @@ def coworking_subject_terms(subject_kind: str, pov_labels: list[str] | None = No
return "the subjects", "them"
def coworking_direction_detail(
def scene_direction_detail(
direction: str,
profile: dict[str, str],
pov_labels: list[str] | None = None,
@@ -145,12 +219,13 @@ def coworking_direction_detail(
foreground = profile["foreground"]
midground = profile["midground"]
background = profile["background"]
detail_label = profile.get("detail_label") or "location details"
subject, pronoun = coworking_subject_terms(subject_kind, pov_labels)
if pov_labels:
if "right side" in direction:
return f"{subject} is in right-side profile; {midground} run behind {pronoun} toward {background}, with coworking details kept at the frame edges"
return f"{subject} is in right-side profile; {midground} run behind {pronoun} toward {background}, with {detail_label} kept at the frame edges"
if "left side" in direction:
return f"{subject} is in left-side profile; {midground} run behind {pronoun} toward {background}, with coworking details kept at the frame edges"
return f"{subject} is in left-side profile; {midground} run behind {pronoun} toward {background}, with {detail_label} kept at the frame edges"
if "back-right" in direction or "back-left" in direction:
return f"{subject} stays close in one continuous diagonal first-person body angle; {midground} lead toward {background} behind {pronoun} at the edges, not in the lower foreground"
if direction == "back view":
@@ -169,7 +244,16 @@ def coworking_direction_detail(
return f"{subject} faces camera beside the {foreground}; {midground} sit between {pronoun} and {background}"
def coworking_distance_detail(
def coworking_direction_detail(
direction: str,
profile: dict[str, str],
pov_labels: list[str] | None = None,
subject_kind: str = "subjects",
) -> str:
return scene_direction_detail(direction, profile, pov_labels, subject_kind)
def scene_distance_detail(
distance: str,
profile: dict[str, str],
subject_kind: str,
@@ -179,18 +263,27 @@ def coworking_distance_detail(
subject, _pronoun = coworking_subject_terms(subject_kind, pov_labels)
if pov_labels:
if "wide" in distance or "full-body" in distance or "full body" in distance:
return f"wide POV keeps {subject} readable with coworking context behind them"
return f"wide POV keeps {subject} readable with {profile['place']} context behind them"
if "close" in distance:
return f"close POV keeps {subject} dominant with coworking context only at the sides or background"
return f"close POV keeps {subject} dominant with {profile['place']} context only at the sides or background"
return f"medium POV keeps {subject} dominant with room context behind them"
if "wide" in distance or "full-body" in distance or "full body" in distance:
return "wide crop keeps floor aisle, table rows, and window depth readable"
return f"wide crop keeps the {profile['foreground']}, {profile['midground']}, and {profile['background']} readable"
if "close" in distance:
return "close crop keeps one desk or counter anchor visible"
return f"close crop keeps one anchor from the {profile['foreground']} visible"
return f"medium crop keeps {subject} dominant"
def coworking_elevation_detail(
def coworking_distance_detail(
distance: str,
profile: dict[str, str],
subject_kind: str,
pov_labels: list[str] | None = None,
) -> str:
return scene_distance_detail(distance, profile, subject_kind, pov_labels)
def scene_elevation_detail(
elevation: str,
profile: dict[str, str],
subject_kind: str,
@@ -200,29 +293,39 @@ def coworking_elevation_detail(
subject, pronoun = coworking_subject_terms(subject_kind, pov_labels)
if pov_labels:
if "low-angle" in elevation:
return f"low angle keeps POV body cues low while windows and partition lines rise behind {pronoun}"
return f"low angle keeps POV body cues low while the {profile['background']} rises behind {pronoun}"
if "elevated" in elevation:
return f"elevated POV keeps the viewer's eye line slightly higher than {subject}, with tabletop and glass lines only behind or at the side edges"
return f"elevated POV keeps the viewer's eye line slightly higher than {subject}, with {profile['foreground']} only behind or at the side edges"
if "high-angle" in elevation:
return f"high angle looks down from the viewer's position with desks and aisle only in the background"
return f"eye-level angle keeps tabletop lines and glass seams behind {pronoun}"
return f"high angle looks down from the viewer's position with {profile['midground']} only in the background"
return f"eye-level angle keeps {profile['midground']} behind {pronoun}"
if "low-angle" in elevation:
return f"low angle keeps the foreground desk edge low while windows and partitions rise behind {pronoun}"
return f"low angle keeps the {profile['foreground']} low while {profile['background']} rises behind {pronoun}"
if "elevated" in elevation:
return f"elevated angle shows tabletop surfaces, laptop shapes, chairs, and walking aisle around {pronoun}"
return f"elevated angle shows the {profile['foreground']} and {profile['midground']} around {pronoun}"
if "high-angle" in elevation:
return f"high angle shows the desk grid, chairs, floor aisle, and placement of {pronoun}"
return f"eye-level angle keeps tabletop lines and glass seams straight"
return f"high angle shows the {profile['place']} layout and placement of {pronoun}"
return f"eye-level angle keeps {profile['midground']} visually stable"
def coworking_camera_scene_directive(
def coworking_elevation_detail(
elevation: str,
profile: dict[str, str],
subject_kind: str,
pov_labels: list[str] | None = None,
) -> str:
return scene_elevation_detail(elevation, profile, subject_kind, pov_labels)
def scene_camera_directive(
scene_text: Any,
parsed: dict[str, Any],
pov_labels: list[str] | None = None,
subject_kind: str = "subjects",
compact_labels: Mapping[str, str] | None = None,
) -> str:
if not is_coworking_scene(scene_text):
profile = scene_camera_profile(scene_text)
if not profile:
return ""
direction = str(parsed.get("orbit_direction") or "").strip()
elevation = str(parsed.get("orbit_elevation_label") or "").strip()
@@ -233,10 +336,9 @@ def coworking_camera_scene_directive(
distance = distance or camera_distance_from_text(custom_prompt)
if not any((direction, elevation, distance, custom_prompt)):
return ""
profile = coworking_location_profile(scene_text)
direction_detail = coworking_direction_detail(direction, profile, pov_labels, subject_kind)
distance_detail = coworking_distance_detail(distance, profile, subject_kind, pov_labels)
elevation_detail = coworking_elevation_detail(elevation, profile, subject_kind, pov_labels)
direction_detail = scene_direction_detail(direction, profile, pov_labels, subject_kind)
distance_detail = scene_distance_detail(distance, profile, subject_kind, pov_labels)
elevation_detail = scene_elevation_detail(elevation, profile, subject_kind, pov_labels)
if pov_labels:
return (
f"{profile['layout_label']} from POV: {direction_detail}. "
@@ -250,19 +352,56 @@ def coworking_camera_scene_directive(
)
def coworking_composition_prompt(scene_text: Any, composition: Any, subject_kind: str = "subjects") -> str:
def coworking_camera_scene_directive(
scene_text: Any,
parsed: dict[str, Any],
pov_labels: list[str] | None = None,
subject_kind: str = "subjects",
compact_labels: Mapping[str, str] | None = None,
) -> str:
if not is_coworking_scene(scene_text):
return ""
return scene_camera_directive(scene_text, parsed, pov_labels, subject_kind, compact_labels)
def profile_composition_text(profile: dict[str, Any], subject_kind: str) -> str:
composition = profile.get("composition") if isinstance(profile.get("composition"), dict) else {}
if subject_kind == "woman" and composition.get("woman"):
return str(composition["woman"])
if subject_kind == "man" and composition.get("man"):
return str(composition["man"])
text = str(composition.get("default") or f"{profile['place']} frame with the subjects clearly placed in the room")
if subject_kind == "couple":
text = text.replace("the subjects", "the couple")
if "composition" not in text.lower():
text = f"{text} composition"
return text
def contextual_composition_prompt(scene_text: Any, composition: Any, subject_kind: str = "subjects") -> str:
text = str(composition or "").strip()
if not text or not is_coworking_scene(scene_text):
if not text:
return text
profile = scene_camera_profile(scene_text)
if not profile:
return text
lower = text.lower()
if not any(term in lower for term in ("office-lobby", "office lobby", "walking composition", "outfit-check")):
profile_lower = " ".join(
str(profile.get(key, "")).lower()
for key in ("place", "foreground", "midground", "background")
)
already_matches = any(term and term in lower for term in profile_lower.replace(",", " ").split())
mismatched = any(term in lower for term in MISMATCHED_COMPOSITION_TERMS)
office_generic = any(term in lower for term in ("office-lobby", "office lobby", "walking composition", "outfit-check"))
if not mismatched and not office_generic and already_matches:
return text
subject, _pronoun = coworking_subject_terms(subject_kind)
if subject_kind == "woman":
return "coworking lounge selfie frame with the woman near a desk edge and tall-window depth behind her"
if subject_kind == "man":
return "coworking lounge portrait frame with the man near a desk edge and tall-window depth behind him"
return f"coworking lounge frame with {subject} near a desk edge and tall-window depth behind them"
if not mismatched and not office_generic and profile.get("family") != "coworking":
return text
return profile_composition_text(profile, subject_kind)
def coworking_composition_prompt(scene_text: Any, composition: Any, subject_kind: str = "subjects") -> str:
return contextual_composition_prompt(scene_text, composition, subject_kind)
def camera_scene_directive_for_context(
@@ -277,7 +416,7 @@ def camera_scene_directive_for_context(
or parsed_camera_config.get("camera_mode") == "disabled"
):
return ""
return coworking_camera_scene_directive(
return scene_camera_directive(
scene_text,
parsed_camera_config,
pov_labels,