This commit is contained in:
mcDandy
2026-01-08 14:14:44 +01:00
parent 01cbdf2e26
commit f273d74ae6
28 changed files with 10849 additions and 2576 deletions
+60 -32
View File
@@ -11,16 +11,17 @@ from .helper_functions import generate_dim_variables, getIndexTensorAlongDim, ev
from .MathNodeBase import MathNodeBase
class VideoMathNode(MathNodeBase):
"""
Enables math expressions on Video (images + audio).
Inputs:
a, b, c, d: Video inputs (b, c, d default to zero if not provided)
w, x, y, z: Float variables for expressions
Audio: Expression for audio component
Images: Expression for image component
Outputs:
VIDEO: Result of applying expressions to input videos
"""
@@ -53,7 +54,7 @@ class VideoMathNode(MathNodeBase):
@classmethod
def execute(cls, Audio, Images, a, b=None, c=None, d=None, w=0.0, x=0.0, y=0.0, z=0.0) -> io.NodeOutput:
ac = a.get_components()
bc = b.get_components() if b is not None else make_zero_like(ac)
cc = c.get_components() if c is not None else make_zero_like(ac)
dc = d.get_components() if d is not None else make_zero_like(ac)
@@ -65,45 +66,72 @@ class VideoMathNode(MathNodeBase):
imgs_d = dc.images.permute(0, 3, 1, 2)
img_vars = {
'a': imgs_a, 'b': imgs_b, 'c': imgs_c, 'd': imgs_d,
'w': w, 'x': x, 'y': y, 'z': z,
'X': getIndexTensorAlongDim(imgs_a, 3),
'Y': getIndexTensorAlongDim(imgs_a, 2),
'B': getIndexTensorAlongDim(imgs_a, 0), 'frame': getIndexTensorAlongDim(imgs_a, 0),
'C': getIndexTensorAlongDim(imgs_a, 1), 'channel': getIndexTensorAlongDim(imgs_a, 1),
'W': imgs_a.shape[3], 'width': imgs_a.shape[3],
'H': imgs_a.shape[2], 'height': imgs_a.shape[2],
'R': float(ac.frame_rate), 'frame_rate': float(ac.frame_rate),
'T': imgs_a.shape[0], 'frame_count': imgs_a.shape[0],
'N': imgs_a.shape[1], 'channel_count': imgs_a.shape[1],
"a": imgs_a,
"b": imgs_b,
"c": imgs_c,
"d": imgs_d,
"w": w,
"x": x,
"y": y,
"z": z,
"X": getIndexTensorAlongDim(imgs_a, 3),
"Y": getIndexTensorAlongDim(imgs_a, 2),
"B": getIndexTensorAlongDim(imgs_a, 0),
"frame": getIndexTensorAlongDim(imgs_a, 0),
"C": getIndexTensorAlongDim(imgs_a, 1),
"channel": getIndexTensorAlongDim(imgs_a, 1),
"W": imgs_a.shape[3],
"width": imgs_a.shape[3],
"H": imgs_a.shape[2],
"height": imgs_a.shape[2],
"R": float(ac.frame_rate),
"frame_rate": float(ac.frame_rate),
"T": imgs_a.shape[0],
"frame_count": imgs_a.shape[0],
"N": imgs_a.shape[1],
"channel_count": imgs_a.shape[1],
} | generate_dim_variables(imgs_a)
result_imgs = eval_tensor_expr(Images, img_vars, imgs_a.shape)
result_imgs = result_imgs.permute(0, 2, 3, 1) # Back to B, H, W, C
# Process audio
audio_a = ac.audio['waveform']
audio_b = bc.audio['waveform']
audio_c = cc.audio['waveform']
audio_d = dc.audio['waveform']
audio_a = ac.audio["waveform"]
audio_b = bc.audio["waveform"]
audio_c = cc.audio["waveform"]
audio_d = dc.audio["waveform"]
audio_vars = {
'a': audio_a, 'b': audio_b, 'c': audio_c, 'd': audio_d,
'w': w, 'x': x, 'y': y, 'z': z,
'B': getIndexTensorAlongDim(audio_a, 0), 'batch': getIndexTensorAlongDim(audio_a, 0),
'C': getIndexTensorAlongDim(audio_a, 1), 'channel': getIndexTensorAlongDim(audio_a, 1),
'S': getIndexTensorAlongDim(audio_a, 2), 'sample': getIndexTensorAlongDim(audio_a, 2),
'R': ac.audio['sample_rate'], 'sample_rate': ac.audio['sample_rate'],
'T': audio_a.shape[2], 'sample_count': audio_a.shape[2],
'N': audio_a.shape[1], 'channel_count': audio_a.shape[1],
"a": audio_a,
"b": audio_b,
"c": audio_c,
"d": audio_d,
"w": w,
"x": x,
"y": y,
"z": z,
"B": getIndexTensorAlongDim(audio_a, 0),
"batch": getIndexTensorAlongDim(audio_a, 0),
"C": getIndexTensorAlongDim(audio_a, 1),
"channel": getIndexTensorAlongDim(audio_a, 1),
"S": getIndexTensorAlongDim(audio_a, 2),
"sample": getIndexTensorAlongDim(audio_a, 2),
"R": ac.audio["sample_rate"],
"sample_rate": ac.audio["sample_rate"],
"T": audio_a.shape[2],
"sample_count": audio_a.shape[2],
"N": audio_a.shape[1],
"channel_count": audio_a.shape[1],
} | generate_dim_variables(audio_a)
result_audio = eval_tensor_expr(Audio, audio_vars, audio_a.shape)
output = VideoFromComponents(VideoComponents(
images=result_imgs,
audio={'waveform': result_audio, 'sample_rate': ac.audio['sample_rate']},
frame_rate=ac.frame_rate,
metadata=ac.metadata
))
output = VideoFromComponents(
VideoComponents(
images=result_imgs,
audio={"waveform": result_audio, "sample_rate": ac.audio["sample_rate"]},
frame_rate=ac.frame_rate,
metadata=ac.metadata,
)
)
return (output,)