Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Variable framerate fix for all branches #1190

Open
wants to merge 1 commit into
base: master
Choose a base branch
from
Open

Variable framerate fix for all branches #1190

wants to merge 1 commit into from

Conversation

qo4on
Copy link

@qo4on qo4on commented Jul 31, 2020

  1. Manim generates variable framerate video instead of the constant value in the CONFIG

  2. mp4fpsmod tool can fix this. Unzip the attachment to manim folder and place this code to line 500 of scene_file_writer.py right after combine_process.wait().

Note: this tool does not work if you have opened the video file in another program.

mp4fpsmod_file_path = os.path.join(
    os.getcwd(),
    'mp4fpsmod.exe' if os.name == 'nt' else 'mp4fpsmod'
)

fix_timestamps_process = subprocess.Popen([
    mp4fpsmod_file_path,
    '--fps', '0:' + str(self.scene.camera.frame_rate),
    '--timescale', '15360',
    movie_file_path, '--inplace'
])
fix_timestamps_process.wait()

Or

mp4fpsmod_file_path = os.path.join(
        os.path.dirname(
            os.path.dirname(
                os.path.dirname(__file__))),
    'mp4fpsmod.exe' if os.name == 'nt' else 'mp4fpsmod'
)

mp4fpsmod.zip

@3b1b
Copy link
Owner

3b1b commented Feb 6, 2021

See this comment.

Still, even if this usage of concat is no longer the default, this might be worth fixing for the concat case in case anyone wants all the animations written to separate files. Is there a way to specify this using ffmpeg, rather than calling mp4fpsmod?

@qo4on
Copy link
Author

qo4on commented Feb 6, 2021

You are right, the separate files have constant frame rate. If they are the goal, you don't need to change anything. For example, if a user concatenates them in Premiere Pro he will get CFR output without any problem.

Is there a way to specify this using ffmpeg, rather than calling mp4fpsmod?

I did not find a way to do this. mp4fpsmod does it in fast lossless mode without any reencoding. ffmpeg can't rearrange frames timestamps losslessly.

ffmpeg is a very buggy thing, it should be used cautiously. Manim also had a problem with wrong colors during png to mp4 conversion. ffmpeg incorrectly defines the output color space. You must add -vf scale=out_color_matrix=bt709 to any rgb to yuv conversion. rgb to yuv always have an error in 8-bit video. But this error should be small: the colors are not exactly the same, they are in a range -2...+2 rgb values. But without -vf scale=out_color_matrix=bt709 the error is around -10...+10.
Also ffmpeg is almost useless when it comes to audio. It changes the waveform and duration of mp3 file when you do mp3 to wav. Many audio filters like alimiter add artifacts...

Another Manim issue I fixed with this wrapper:

def time_cor(run_time, time_shift=0, frame_time=1 / 60):
    whole, rem = divmod(run_time, frame_time)
    if rem + time_shift > frame_time / 2:
        run_time = (whole + 1) * frame_time
        time_shift += rem - frame_time
    else:
        run_time = whole * frame_time
        time_shift += rem
    return run_time, time_shift

I wrapped every run_time with this function to fix "rounding to frame" time error. Imagine you have 1000 small partial video clips and run_time of each of them cannot be divided by fps without a reminder. This means you will get a final video a few seconds longer than expected.

eulertour pushed a commit to eulertour/manim-3b1b that referenced this pull request Mar 31, 2021
@RadicalGDPRist
Copy link

Doesn't this tool only change the PTS and DTS? There's -bsf:v 'setts=dts... in ffmpeg as argued based on and over here

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

3 participants