Update app.py

app.py

@@ -65,6 +65,13 @@ def apply_linear_gain(audio_clip, gain_linear: float):
 # ---------- Image utilities ----------
 
 def load_and_fit_image(path: str, width: int, height: int, fit: str = "contain", bg: str = "#000000") -> np.ndarray:
+    """
+    Loads an image file and returns an RGB numpy array with exact (height, width, 3).
+    fit modes:
+      - "contain": letterbox to fit within target size (keeps aspect), background color fills the rest.
+      - "cover":   fill target size (keeps aspect) with center crop.
+      - "stretch": distort to target size.
+    """
     img = Image.open(path).convert("RGB")
 
     if fit == "stretch":
@@ -76,19 +83,22 @@ def load_and_fit_image(path: str, width: int, height: int, fit: str = "contain",
     src_aspect = float(iw) / float(ih)
 
     if fit == "cover":
-        if src_aspect > target_asect := target_aspect:
+        # scale to cover, then center-crop
+        if src_aspect > target_aspect:
+            # image too wide -> fit height, crop width
             new_h = height
-            new_w = int(round(src_asect * new_h))
+            new_w = int(round(src_aspect * new_h))
         else:
+            # image too tall -> fit width, crop height
             new_w = width
-            new_h = int(round(new_w / src_asect))
+            new_h = int(round(new_w / src_aspect))
         img = img.resize((new_w, new_h), Image.LANCZOS)
         left = (new_w - width) // 2
         top = (new_h - height) // 2
         img = img.crop((left, top, left + width, top + height))
         return np.array(img)
 
-    # contain
+    # contain (letterbox/pillarbox)
     canvas = Image.new("RGB", (width, height), bg)
     if src_aspect > target_aspect:
         new_w = width
@@ -103,6 +113,7 @@ def load_and_fit_image(path: str, width: int, height: int, fit: str = "contain",
     return np.array(canvas)
 
 
+
 # ---------- TTS backends ----------
 
 _TTS_CACHE: Dict[str, object] = {}