Speed up screenshot OCR by downscaling before tesseract

app/screenshot.py

@@ -51,16 +51,28 @@ SECOND_PASS_MIN_SHRINK = 0.02
 # OCR via tesseract subprocess
 # ──────────────────────────────────────────────────────────────
 
+OCR_MAX_DIM = 1500  # downscale before tesseract for speed; boxes scaled back
+
+
 def run_tesseract(image: np.ndarray, min_conf: int = 30) -> list[tuple]:
     """Call `tesseract` CLI, parse TSV output, return (x, y, w, h) boxes."""
+    h, w = image.shape[:2]
+    scale = 1.0
+    ocr_image = image
+    if max(h, w) > OCR_MAX_DIM:
+        scale = OCR_MAX_DIM / float(max(h, w))
+        new_w = max(1, int(round(w * scale)))
+        new_h = max(1, int(round(h * scale)))
+        ocr_image = cv2.resize(image, (new_w, new_h), interpolation=cv2.INTER_AREA)
+
     tmp = tempfile.NamedTemporaryFile(suffix=".png", delete=False)
     try:
-        Image.fromarray(image).save(tmp.name)
+        Image.fromarray(ocr_image).save(tmp.name)
         result = subprocess.run(
             ["tesseract", tmp.name, "stdout", "--psm", "3", "tsv"],
             capture_output=True,
             text=True,
-            timeout=30,
+            timeout=60,
         )
     except FileNotFoundError:
         print("[screenshot] tesseract binary not found")
@@ -105,12 +117,21 @@ def run_tesseract(image: np.ndarray, min_conf: int = 30) -> list[tuple]:
             continue
         if conf < min_conf:
             continue
-        boxes.append((
-            int(cols[idx_left]),
-            int(cols[idx_top]),
-            int(cols[idx_width]),
-            int(cols[idx_height]),
-        ))
+        if scale != 1.0:
+            inv = 1.0 / scale
+            boxes.append((
+                int(round(int(cols[idx_left]) * inv)),
+                int(round(int(cols[idx_top]) * inv)),
+                int(round(int(cols[idx_width]) * inv)),
+                int(round(int(cols[idx_height]) * inv)),
+            ))
+        else:
+            boxes.append((
+                int(cols[idx_left]),
+                int(cols[idx_top]),
+                int(cols[idx_width]),
+                int(cols[idx_height]),
+            ))
     return boxes