Fix issue 2269 - And a lot of other things

[OsaAjani]

TL;DR: Transparency was broken all around, I fixed it.

What started as a "quick fix" for issue #2269 rapidly escalated into (re)discovery (see #2012 (comment)) of multiple bugs in regards to transparency handling in moviepy. To summarize, transparency was plain and simple broken :

• Reading video with transparency didn't work
• Writing video with background transparency didn't work
• Calculating final transparency of composite clip when the final clip should be partly transparent didn't work
• Using composite clip of composite clips whom add some clips with mask didn't work (original MoviePy2: TextClip transperancy glitch with multiple CompositeVideoClip #2269)
• Compositing and bliting was partially buggy

Basically the only part that was working as intended was composite clip with basic clips inside, and a fully opaque background.

I also added codec extraction to FFMPEGParser when possible, as this info was required for proper transparency handling on ffmpeg.

I think this should fix not only #2269, but also #2109, #2082, #2008, #1978, #1935, #1881, #1821, #1759, #1301, #896, #400, and probably others I just dont know about...

• I have provided code that clearly demonstrates the bug and that only works correctly when applying this fix
• I have added suitable tests demonstrating a fixed bug or new/changed feature to the test suite in tests/
• I have properly documented new or changed features in the documentation or in the docstrings
• I have properly explained unusual or unexpected code in the comments around it

[OsaAjani]

I will give it a few days for other to look at the PR and flag anything they think is wrong, then if not protest will proceed with merging !

[OsaAjani]

Hey @Zulko @keikoro and others, if someone want to take a look at this PR to check I didn't do anything stupid I intend to merge it during the next two days without responses :)

[claell]

@OsaAjani will this also fix #195 (and incorporate some solution as was apparently proposed in #1395)?

[Implosiv3]

Print here.

[Implosiv3]

In new line 302, there is a bug. It should be mask = clip.mask or ColorClip(clip.size, color=1, is_mask=True) instead of the [1, 1] mask size.

You can see the issue here (#2247) and my PR here (#2262).

[OsaAjani]

Thanks, I will fix this and include the changes in your PR in this one, as well as adding a unit test

[OsaAjani]

I'll publish with a new tag 2.1.2 tomorrow

[claell]

@OsaAjani It seems that the release takes a little longer. Can you share a new (rough) ETA?

[OsaAjani]

Sorry, last days where more about home fixing than coding ^^. Definitely in the week.

[claell]

No problem, thanks for the update! Was just checking frequently the past days to see whether it already got released. 😄

[OsaAjani]

Done.

[claell]

Thanks! Looks as if the publish to PyPI failed, though? The logs indicate that the file already existed, or something like that.

[OsaAjani]

I fixed it

[claell]

Awesome! Just updated with pip and can confirm that the transparency problems in #2269 are fixed now, for me!

[OsaAjani]

@claell actually after a few tests I think this PR mostly solved the aliasing problem with text, at least judging by this example:

from moviepy import *
import numpy as np


vclip = VideoFileClip('../resources/bbb.mp4').subclipped(10, 11)

tClip = TextClip(duration=3, font="../resources/font/example.ttf", text="Some string", color="red", font_size=250, bg_color="black")
tClip.show(1)

tClip2 = TextClip(duration=3, font="../resources/font/example.ttf", text="Some string", color="red", font_size=250)
bg = ColorClip(tClip2.size, color=(255, 255, 255))
print(np.unique(tClip2.mask.get_frame(1)))
cClip = CompositeVideoClip([vclip, tClip2, tClip2.with_position(("center", "center"))], use_bgclip=True)
cClip.show(1)