Reject incompatible overhead silhouettes

app.py

@@ -1947,8 +1947,8 @@ def neural_foreground_cutout(image: Image.Image) -> Image.Image:
     return cutout
 
 
-def top_down_subject_contract(spec: AssetSpec) -> str:
-    """Describe visible overhead anatomy without drawing or reshaping model artwork."""
+def infer_top_down_subject_geometry(spec: AssetSpec) -> str:
+    """Infer an overhead footprint from user-authored visual terms, never the asset ID."""
     # Classify only from the user's visual description. Custom IDs such as
     # ``player`` or ``enemy`` are integration hooks, not reliable anatomy.
     description = f"{spec.description} {spec.prompt}".lower()
@@ -1985,16 +1985,28 @@ def top_down_subject_contract(spec: AssetSpec) -> str:
     )
 
     if contains_any_term(description, radial_terms):
+        return "radial"
+    if contains_any_term(description, vehicle_terms):
+        return "vehicle"
+    if spec.silhouette == "humanoid" or contains_any_term(description, humanoid_terms):
+        return "humanoid"
+    return "generic"
+
+
+def top_down_subject_contract(spec: AssetSpec) -> str:
+    """Describe visible overhead anatomy without drawing or reshaping model artwork."""
+    geometry = infer_top_down_subject_geometry(spec)
+    if geometry == "radial":
         return (
             "Top of body dominates; splayed limbs radiate in image plane; "
             "no upright torso or side elevation."
         )
-    if contains_any_term(description, vehicle_terms):
+    if geometry == "vehicle":
         return (
             "Roof and chassis top dominate; wheel footprints flank both sides; "
             "no grille or side elevation."
         )
-    if spec.silhouette == "humanoid" or contains_any_term(description, humanoid_terms):
+    if geometry == "humanoid":
         return (
             "Full body overhead: head crown, shoulders and back visible; "
             "face and chest invisible; limbs foreshortened."
@@ -2120,7 +2132,7 @@ def primary_diffusion_png(spec: AssetSpec, index: int, run_id: int) -> tuple[byt
                 continue
             if has_front_view_like_top_down_silhouette(content, spec):
                 last_failure_detail = (
-                    "the last output retained a tall portrait-like silhouette instead of the requested overhead view"
+                    "the last output silhouette conflicted with the requested overhead body footprint"
                 )
                 continue
             if has_undersized_foreground_subject(content):
@@ -2467,13 +2479,13 @@ def has_implausibly_thin_foreground_subject(content: bytes, spec: AssetSpec) ->
 
 
 def has_front_view_like_top_down_silhouette(content: bytes, spec: AssetSpec) -> bool:
-    """Conservatively reject upright portrait silhouettes for overhead sprite contracts.
+    """Reject silhouettes that contradict a requested overhead body footprint.
 
-    Alpha geometry cannot prove camera perspective, but a tall, canvas-filling
-    humanoid is a strong signal that the generator ignored an overhead request.
-    Broader non-humanoid checks apply only when the user's own description asks
-    for a radial, wide, or multi-legged body. Narrow top-down props and vehicles
-    therefore remain valid user-configurable assets.
+    Alpha geometry cannot prove camera perspective. It can, however, fail closed
+    when a normalized, canvas-filling silhouette contradicts explicit user terms:
+    compact humanoids cannot remain upright portraits; radial bodies must not be
+    taller than wide; and explicitly wide vehicles must present a wide footprint.
+    Narrow props remain exempt.
     """
     if spec.camera != "top_down" or spec.composition not in {"single_subject", "icon", "animation_frame"}:
         return False
@@ -2491,31 +2503,19 @@ def has_front_view_like_top_down_silhouette(content: bytes, spec: AssetSpec) ->
     if height_ratio < 0.72:
         return False
 
-    if spec.silhouette == "humanoid" and aspect >= 1.35:
+    geometry = infer_top_down_subject_geometry(spec)
+    if geometry == "humanoid" and aspect >= 1.10:
+        return True
+
+    if geometry == "radial" and aspect >= 1.02:
         return True
 
     description = f"{spec.description} {spec.prompt}"
-    broad_shape_requested = contains_any_term(
+    explicitly_wide_vehicle = geometry == "vehicle" and contains_any_term(
         description,
-        (
-            "broad",
-            "radial",
-            "wide-bodied",
-            "wide bodied",
-            "six-legged",
-            "six legged",
-            "four-legged",
-            "four legged",
-            "multi-legged",
-            "multi legged",
-            "quadruped",
-            "spider-like",
-            "spider like",
-            "crab-like",
-            "crab like",
-        ),
+        ("wide", "four-wheel", "four wheel", "six-wheel", "six wheel", "tracked"),
     )
-    if broad_shape_requested and aspect >= 1.45:
+    if explicitly_wide_vehicle and aspect >= 0.95:
         return True
 
     narrow_subject_requested = contains_any_term(
@@ -2840,7 +2840,7 @@ def validate_asset_png(content: bytes, spec: AssetSpec) -> list[str]:
             elif spec.expected_subjects == 1 and has_implausibly_thin_foreground_subject(content, spec):
                 warnings.append("sprite foreground silhouette is implausibly thin")
             elif spec.expected_subjects == 1 and has_front_view_like_top_down_silhouette(content, spec):
-                warnings.append("top-down sprite silhouette remains strongly portrait-like")
+                warnings.append("top-down sprite silhouette conflicts with the requested overhead footprint")
             elif spec.expected_subjects == 1 and has_undersized_foreground_subject(content):
                 warnings.append("foreground subject occupies too little of the output canvas")
     if spec.composition == "seamless" and image.width > 1 and image.height > 1: