Extract scene camera adapters
This commit is contained in:
@@ -0,0 +1,286 @@
|
||||
from __future__ import annotations
|
||||
|
||||
from typing import Any, Mapping
|
||||
|
||||
|
||||
CAMERA_DIRECTIONS = (
|
||||
"front-right quarter view",
|
||||
"right side view",
|
||||
"back-right quarter view",
|
||||
"back view",
|
||||
"back-left quarter view",
|
||||
"left side view",
|
||||
"front-left quarter view",
|
||||
"front view",
|
||||
)
|
||||
|
||||
CAMERA_ELEVATIONS = ("low-angle shot", "eye-level shot", "elevated shot", "high-angle shot")
|
||||
CAMERA_DISTANCES = (
|
||||
"wide shot",
|
||||
"full-body shot",
|
||||
"three-quarter body shot",
|
||||
"medium shot",
|
||||
"close-up",
|
||||
"extreme close-up",
|
||||
)
|
||||
|
||||
|
||||
def is_coworking_scene(scene_text: Any) -> bool:
|
||||
text = str(scene_text or "").lower()
|
||||
return any(
|
||||
term in text
|
||||
for term in (
|
||||
"coworking",
|
||||
"cowork",
|
||||
"office lounge",
|
||||
"business cafe",
|
||||
"work cafe",
|
||||
"shared office",
|
||||
"corporate office",
|
||||
"office after hours",
|
||||
"laptops",
|
||||
"warm desks",
|
||||
"repeating desks",
|
||||
"glass partitions",
|
||||
"copier alcove",
|
||||
)
|
||||
)
|
||||
|
||||
|
||||
def _compact_label(value: Any, compact_labels: Mapping[str, str] | None = None) -> str:
|
||||
text = str(value or "")
|
||||
if compact_labels and text in compact_labels:
|
||||
return compact_labels[text]
|
||||
return text.replace("_", " ")
|
||||
|
||||
|
||||
def camera_geometry_phrase(parsed: dict[str, Any], compact_labels: Mapping[str, str] | None = None) -> str:
|
||||
direction = str(parsed.get("orbit_direction") or "").strip()
|
||||
elevation = str(parsed.get("orbit_elevation_label") or "").strip()
|
||||
distance = str(parsed.get("orbit_distance_label") or "").strip()
|
||||
custom = str(parsed.get("custom_camera_prompt") or "").strip()
|
||||
if not any((direction, elevation, distance)) and custom:
|
||||
return custom
|
||||
parts = [part for part in (direction, elevation, distance) if part and part != "auto"]
|
||||
if parts:
|
||||
return ", ".join(parts)
|
||||
compact_parts = [
|
||||
_compact_label(parsed.get(key), compact_labels)
|
||||
for key in ("shot_size", "angle", "distance")
|
||||
]
|
||||
compact_parts = [part for part in compact_parts if part and part != "auto"]
|
||||
return ", ".join(compact_parts)
|
||||
|
||||
|
||||
def camera_direction_from_text(text: Any) -> str:
|
||||
source = str(text or "").lower()
|
||||
for label in CAMERA_DIRECTIONS:
|
||||
if label in source:
|
||||
return label
|
||||
return ""
|
||||
|
||||
|
||||
def camera_elevation_from_text(text: Any) -> str:
|
||||
source = str(text or "").lower()
|
||||
for label in CAMERA_ELEVATIONS:
|
||||
if label in source:
|
||||
return label
|
||||
return ""
|
||||
|
||||
|
||||
def camera_distance_from_text(text: Any) -> str:
|
||||
source = str(text or "").lower()
|
||||
for label in CAMERA_DISTANCES:
|
||||
if label in source:
|
||||
return label
|
||||
return ""
|
||||
|
||||
|
||||
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",
|
||||
}
|
||||
|
||||
|
||||
def coworking_subject_terms(subject_kind: str, pov_labels: list[str] | None = None) -> tuple[str, str]:
|
||||
if pov_labels:
|
||||
return "the visible partner", "them"
|
||||
if subject_kind == "woman":
|
||||
return "the woman", "her"
|
||||
if subject_kind == "man":
|
||||
return "the man", "him"
|
||||
if subject_kind == "couple":
|
||||
return "the couple", "them"
|
||||
return "the subjects", "them"
|
||||
|
||||
|
||||
def coworking_direction_detail(
|
||||
direction: str,
|
||||
profile: dict[str, str],
|
||||
pov_labels: list[str] | None = None,
|
||||
subject_kind: str = "subjects",
|
||||
) -> str:
|
||||
direction = str(direction or "").strip().lower()
|
||||
foreground = profile["foreground"]
|
||||
midground = profile["midground"]
|
||||
background = profile["background"]
|
||||
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"
|
||||
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"
|
||||
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":
|
||||
return f"the viewer looks past {subject}'s back toward {midground}, then into {background}; only POV body cues sit low in frame"
|
||||
if "front-right" in direction or "front-left" in direction:
|
||||
return f"{subject} fills the first-person front-quarter view; {midground} recede diagonally behind {pronoun} toward {background}"
|
||||
return f"{subject} faces the viewer in first-person view; {midground} and {background} stay behind {pronoun}, not between viewer and body"
|
||||
if "right side" in direction or "left side" in direction:
|
||||
return f"{subject} is held in side profile along the {foreground}; {midground} run laterally behind {pronoun}, with {background} still readable"
|
||||
if "back-right" in direction or "back-left" in direction:
|
||||
return f"{subject} is viewed from a rear-quarter angle, partly turning back toward camera; the {foreground} stays low in frame while {midground} lead into {background}"
|
||||
if direction == "back view":
|
||||
return f"{subject} is seen from behind with the {foreground} at camera side, facing into {midground} and {background}"
|
||||
if "front-right" in direction or "front-left" in direction:
|
||||
return f"{subject} is placed beside the {foreground}; {midground} recede diagonally behind {pronoun} toward {background}"
|
||||
return f"{subject} faces camera beside the {foreground}; {midground} sit between {pronoun} and {background}"
|
||||
|
||||
|
||||
def coworking_distance_detail(
|
||||
distance: str,
|
||||
profile: dict[str, str],
|
||||
subject_kind: str,
|
||||
pov_labels: list[str] | None = None,
|
||||
) -> str:
|
||||
distance = str(distance or "").strip().lower()
|
||||
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"
|
||||
if "close" in distance:
|
||||
return f"close POV keeps {subject} dominant with coworking 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"
|
||||
if "close" in distance:
|
||||
return "close crop keeps one desk or counter anchor visible"
|
||||
return f"medium crop keeps {subject} dominant"
|
||||
|
||||
|
||||
def coworking_elevation_detail(
|
||||
elevation: str,
|
||||
profile: dict[str, str],
|
||||
subject_kind: str,
|
||||
pov_labels: list[str] | None = None,
|
||||
) -> str:
|
||||
elevation = str(elevation or "").strip().lower()
|
||||
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}"
|
||||
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"
|
||||
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}"
|
||||
if "low-angle" in elevation:
|
||||
return f"low angle keeps the foreground desk edge low while windows and partitions rise behind {pronoun}"
|
||||
if "elevated" in elevation:
|
||||
return f"elevated angle shows tabletop surfaces, laptop shapes, chairs, and walking aisle 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"
|
||||
|
||||
|
||||
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 ""
|
||||
direction = str(parsed.get("orbit_direction") or "").strip()
|
||||
elevation = str(parsed.get("orbit_elevation_label") or "").strip()
|
||||
distance = str(parsed.get("orbit_distance_label") or "").strip()
|
||||
custom_prompt = str(parsed.get("custom_camera_prompt") or "").strip()
|
||||
direction = direction or camera_direction_from_text(custom_prompt)
|
||||
elevation = elevation or camera_elevation_from_text(custom_prompt)
|
||||
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)
|
||||
if pov_labels:
|
||||
return (
|
||||
f"{profile['layout_label']} from POV: {direction_detail}. "
|
||||
f"{distance_detail}; {elevation_detail}; use the multiangle camera only as first-person spatial geometry."
|
||||
)
|
||||
geometry = camera_geometry_phrase(parsed, compact_labels)
|
||||
geometry_clause = f" ({geometry})" if geometry else ""
|
||||
return (
|
||||
f"{profile['layout_label']}{geometry_clause}: {direction_detail}; "
|
||||
f"{distance_detail}; {elevation_detail}."
|
||||
)
|
||||
|
||||
|
||||
def coworking_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):
|
||||
return text
|
||||
lower = text.lower()
|
||||
if not any(term in lower for term in ("office-lobby", "office lobby", "walking composition", "outfit-check")):
|
||||
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"
|
||||
|
||||
|
||||
def camera_scene_directive_for_context(
|
||||
scene_text: Any,
|
||||
parsed_camera_config: dict[str, Any],
|
||||
pov_labels: list[str] | None = None,
|
||||
subject_kind: str = "subjects",
|
||||
compact_labels: Mapping[str, str] | None = None,
|
||||
) -> str:
|
||||
if (
|
||||
parsed_camera_config.get("camera_detail") == "off"
|
||||
or parsed_camera_config.get("camera_mode") == "disabled"
|
||||
):
|
||||
return ""
|
||||
return coworking_camera_scene_directive(
|
||||
scene_text,
|
||||
parsed_camera_config,
|
||||
pov_labels,
|
||||
subject_kind,
|
||||
compact_labels,
|
||||
)
|
||||
Reference in New Issue
Block a user