Skip to content

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.

Fix nondeterministic video decode at unaligned widths (CORE-299) (#14438)

John Pollocka1d95f3f8266
diff --git a/comfy_api/latest/_input_impl/video_types.py b/comfy_api/latest/_input_impl/video_types.py
index dfdf58515..92a1298c0 100644
--- a/comfy_api/latest/_input_impl/video_types.py
+++ b/comfy_api/latest/_input_impl/video_types.py
@@ -270,6 +270,7 @@ class VideoFromFile(VideoInput):
image_format = 'gbrpf32le'
process_image_format = lambda a: a
+ align_graph = None
audio = None
streams = [video_stream]
@@ -323,7 +324,24 @@ class VideoFromFile(VideoInput):
checked_alpha = True
- img = frame.to_ndarray(format=image_format) # shape: (H, W, 4)
+ # Fix non-deterministic video decode when the video width is not a multiple of 32
+ # For non-yuvj pixel formats (all H.264/H.265 video)
+ if image_format in ('gbrpf32le', 'gbrapf32le') and frame.width % 32 != 0:
+ if align_graph is None:
+ pad_w = ((frame.width + 31) // 32) * 32
+ g = av.filter.Graph()
+ g_src = g.add_buffer(width=frame.width, height=frame.height,
+ format=frame.format.name, time_base=video_stream.time_base)
+ g_pad = g.add('pad', f'{pad_w}:{frame.height}:0:0')
+ g_sink = g.add('buffersink')
+ g_src.link_to(g_pad)
+ g_pad.link_to(g_sink)
+ g.configure()
+ align_graph = (g, g_src, g_sink)
+ align_graph[1].push(frame)
+ img = np.ascontiguousarray(align_graph[2].pull().to_ndarray(format=image_format)[:, :frame.width])
+ else:
+ img = frame.to_ndarray(format=image_format)
if frame.rotation != 0:
k = int(round(frame.rotation // 90))
img = np.rot90(img, k=k, axes=(0, 1)).copy()