diff.py

""" Diff video sources to catch infoboxes and other signs.

Usage: $ python diff.py <clean> <hardsubbed> [start-offset]

`start-offset` should be the number of frames to drop from start to align the two video sources.
Positive values drop frames from `clean`, negative from `hardsubbed`.
If left out, automatically deduce the proper offset.

Outputs frame ranges to stdout, and a diffing clip to "out.y4m".

Requires vapoursynth and lvsfunc, both available from PyPi (shouldn't require others? idk)
"""

import vapoursynth as vs
import lvsfunc
import sys

core = vs.core

print(f"!! args: {sys.argv}")
if len(sys.argv) < 3:
    raise ValueError("Give at least the two filenames!")
    sys.exit(1)
fclean, fsigns, sync, *_ = sys.argv[1:] + [None]

clean = lvsfunc.src(sys.argv[1])
signs = lvsfunc.src(sys.argv[2])

# verify that sources line up...
# assume the only thing to sync is removing frames at start

if sync is None:
    # try to align automatically
    sync = len(clean) - len(signs)
    print(f"!! automatically syncing: {sync} ({len(clean)} - {len(signs)})")
if sync < 0:
    signs = signs[-sync:]
else:
    clean = clean[sync:]

# fix resolution differences, if applicable, by rescaling to a common resolution
clean = clean.resize.Bicubic(signs.width, signs.height, signs.format)

print(clean)
print(signs)

# tweak `thr` to adjust threshold. higher catches smaller diffs
# 72 should catch everything, but going up to ~96 can help (but will start to catch artifacts)
comp, ranges = lvsfunc.comparison.diff(clean, signs, return_ranges=True, thr=72)

print("differences on ranges: {}".format(", ".join(f"{a}-{b}" for a,b in ranges)))
# comp.set_output()
with open('out.y4m', 'wb') as fo:
    comp.output(fo, y4m=True)