refactor: modernize type hinting by using collection ABCs and built-in generic types across the codebase

app.py

@@ -315,9 +315,7 @@ def on_video_upload(video_path: str | None):
 
 def on_preview_crop(editor_value: dict | None, meta_state: dict | None, context_px: int):
     """Compute crop region from mask and render a preview overlay."""
-    if meta_state is None:
-        return gr.update(), "Upload a video first."
-    if editor_value is None:
+    if meta_state is None or editor_value is None:
         return gr.update(), "Upload a video first."
 
     raw_mask = _get_mask_from_editor(editor_value)
@@ -333,7 +331,7 @@ def on_preview_crop(editor_value: dict | None, meta_state: dict | None, context_
         if bg is None:
             first_frame = np.zeros((meta.height, meta.width, 3), dtype=np.uint8)
         else:
-            first_frame = np.array(Image.fromarray(np.array(bg)).convert("RGB"))
+            first_frame = np.array(Image.fromarray(np.asarray(bg)).convert("RGB"))
             # Ensure first_frame is at full source resolution for the overlay
             if first_frame.shape[1] != meta.width or first_frame.shape[0] != meta.height:
                 first_frame = np.array(

pipeline/composite.py

@@ -22,7 +22,7 @@ exactly once and reuses the result for every frame's ``composite_with_alpha``.
 from __future__ import annotations
 
 from pathlib import Path
-from typing import Iterable
+from collections.abc import Iterable
 
 import numpy as np
 from PIL import Image

pipeline/crop.py

@@ -18,7 +18,6 @@ from __future__ import annotations
 
 import math
 from dataclasses import dataclass
-from typing import Tuple
 
 import numpy as np
 from scipy.ndimage import binary_dilation
@@ -182,7 +181,7 @@ class CropRegion:
         return self.frame_y + self.frame_h
 
     @property
-    def pil_box(self) -> Tuple[int, int, int, int]:
+    def pil_box(self) -> tuple[int, int, int, int]:
         """(left, upper, right, lower) — 4-tuple expected by PIL.Image.crop()."""
         return (self.frame_x, self.frame_y, self.frame_x2, self.frame_y2)
 
@@ -252,7 +251,7 @@ def mask_to_bbox(mask: np.ndarray) -> BBox:
 def find_target_resolution(
     required_w: int,
     required_h: int,
-) -> Tuple[int, int]:
+) -> tuple[int, int]:
     """
     Find the smallest VACE-compatible (width, height) pair such that
     width >= required_w AND height >= required_h.
@@ -489,7 +488,7 @@ def _centre_crop(
     crop_h: int,
     frame_w: int,
     frame_h: int,
-) -> Tuple[int, int]:
+) -> tuple[int, int]:
     """
     Compute top-left (x, y) of a crop_w x crop_h rectangle centred at
     (cx, cy), then clamp so the rectangle stays within the frame.

pipeline/lama.py

@@ -23,7 +23,7 @@ from __future__ import annotations
 
 import os
 from pathlib import Path
-from typing import Generator, List
+from collections.abc import Generator
 
 import numpy as np
 from PIL import Image
@@ -93,7 +93,7 @@ def _get_model():
 # ---------------------------------------------------------------------------
 
 def inpaint_frames_lama_stream(
-    frame_paths: List[Path],
+    frame_paths: list[Path],
     crop_region: CropRegion,
     inpaint_mask: np.ndarray,
     progress_fn=None,

pipeline/vace.py

@@ -61,7 +61,8 @@ from __future__ import annotations
 import os
 import threading
 from pathlib import Path
-from typing import Callable, Generator, List, Optional
+from collections.abc import Callable, Generator
+from typing import Optional
 
 import numpy as np
 import torch
@@ -310,16 +311,16 @@ def _get_pipe():
 # ---------------------------------------------------------------------------
 
 def _load_chunk_crops(
-    frame_paths: List[Path],
+    frame_paths: list[Path],
     crop_region: CropRegion,
     target_w: int,
     target_h: int,
     start: int,
     end: int,
-) -> List[Image.Image]:
+) -> list[Image.Image]:
     """Load crops for [start, end) at VACE target resolution."""
     box = crop_region.pil_box
-    crops: List[Image.Image] = []
+    crops: list[Image.Image] = []
     for fp in frame_paths[start:end]:
         img = Image.open(fp).convert("RGB")
         crops.append(img.crop(box).resize((target_w, target_h), Image.LANCZOS))
@@ -331,7 +332,7 @@ def _load_chunk_crops(
 # ---------------------------------------------------------------------------
 
 def inpaint_frames_vace_stream(
-    frame_paths: List[Path],
+    frame_paths: list[Path],
     crop_region: CropRegion,
     inpaint_mask: np.ndarray,
     num_inference_steps: Optional[int] = None,
@@ -457,5 +458,3 @@ def inpaint_frames_vace_stream(
         del result
         if torch.cuda.is_available():
             torch.cuda.empty_cache()
-
-

pipeline/video.py

@@ -24,7 +24,7 @@ import subprocess
 import tempfile
 from dataclasses import dataclass
 from pathlib import Path
-from typing import List, Optional
+from typing import Optional
 
 import numpy as np
 from PIL import Image
@@ -167,7 +167,7 @@ def extract_frames(
     pattern: str = "%06d.png",
     fps: float | None = None,
     max_duration_s: float | None = None,
-) -> List[Path]:
+) -> list[Path]:
     """
     Extract every frame from *video_path* as PNG images into *out_dir*.