fix: reject sub-32px frames in compute_crop_region

The VACE / Wan2.1 VAE downsamples spatially by 32×, so each frame
dimension must be at least 32px. The existing crop logic floored the
target to a multiple of 32 then forced a min of 32 — but on a 30×30
frame, that produced a target larger than the frame, silently
overflowing downstream numpy slices in pipeline.composite.

Now compute_crop_region rejects frames smaller than MIN_FRAME_DIM=32
on either axis with a clean ValueError, surfacing the failure at the
point of detection instead of as a confusing slice-shape mismatch
later in the pipeline.

Reachable in practice via accidentally-tiny inputs (thumbnails uploaded
as videos, etc.).

Found by tests/test_callbacks.py round.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

pipeline/crop.py

@@ -42,6 +42,13 @@ MAX_WATERMARK_AREA_FRACTION = 0.5
 # drawing so soft watermark edges don't bleed through after compositing.
 DEFAULT_DILATE_PX = 6
 
+# VACE / Wan2.1 VAE downsamples spatially by 32×, so each frame dimension
+# must be at least 32px. Frames smaller than this are unprocessable; the
+# floor-to-32 + max(32) logic in compute_crop_region would otherwise
+# silently emit a target larger than the frame, overflowing downstream
+# numpy slices in pipeline.composite.
+MIN_FRAME_DIM = 32
+
 
 # ---------------------------------------------------------------------------
 # Supported VACE / Wan2.1 resolutions
@@ -336,6 +343,11 @@ def compute_crop_region(
     ValueError
         If the watermark bbox is larger than the full frame (sanity check).
     """
+    if frame_w < MIN_FRAME_DIM or frame_h < MIN_FRAME_DIM:
+        raise ValueError(
+            f"Frame is too small ({frame_w}×{frame_h}). "
+            f"Minimum supported dimension is {MIN_FRAME_DIM}px on each side."
+        )
     if watermark_bbox.width > frame_w or watermark_bbox.height > frame_h:
         raise ValueError(
             f"Watermark bbox {watermark_bbox} is larger than the frame "