Add avfilter_process_command

[caffeinism]

I'm writing a PR for this issue. I think LLMs are dumb, but I think I'm dumber when it comes to Cython and libav, so please understand that I wrote the code as a dumb-squared combination. I really want to use this feature and hope you'll consider this a draft.

import av
import av.filter
import numpy as np

w = 1024
h = 1024

x = np.linspace(0, 255, w, dtype=np.uint8)
y = np.linspace(0, 255, h, dtype=np.uint8)
arr = np.zeros((h, w, 3), dtype=np.uint8)
arr[..., 0] = x[None, :]
arr[..., 1] = y[:, None]
arr[..., 2] = 128
frame = av.VideoFrame.from_ndarray(arr, format="rgb24")

graph = av.filter.Graph()

src = graph.add_buffer(template=frame)
scale = graph.add("scale", args=f"w={512}:h={512}")
sink = graph.add("buffersink")

src.link_to(scale)
scale.link_to(sink)
graph.configure()

src.push(frame)
out1 = sink.pull()
print(f"1st: {out1.width}x{out1.height}")

scale.process_command("width", "512")
scale.process_command("height", "256")

src.push(frame)
out2 = sink.pull()
print(f"2nd: {out2.width}x{out2.height}")

frame:

out1:

out2:

import av
import av.filter
import numpy as np

w = 1024
h = 1024

x = np.linspace(0, 255, w, dtype=np.uint8)
y = np.linspace(0, 255, h, dtype=np.uint8)
arr = np.zeros((h, w, 3), dtype=np.uint8)
arr[..., 0] = x[None, :]
arr[..., 1] = y[:, None]
arr[..., 2] = 128
frame = av.VideoFrame.from_ndarray(arr, format="rgb24")
square = av.VideoFrame.from_ndarray(np.zeros((256, 256, 3), dtype=np.uint8), format="rgb24")
frame.pts = square.pts = None

graph = av.filter.Graph()

src = graph.add_buffer(template=frame)
dst = graph.add_buffer(template=square)
overlay = graph.add("overlay", args=f"x={256}:y={256}")
sink = graph.add("buffersink")

src.link_to(overlay, input_idx=0)
dst.link_to(overlay, input_idx=1)
overlay.link_to(sink)
graph.configure()


src.push(frame)
dst.push(square)
out1 = sink.pull()

overlay.process_command("x", "512")
overlay.process_command("y", "512")

src.push(frame)
dst.push(square)
out2 = sink.pull()

out1

out2