Hey Artur,
i can confirm the behaviour you describe, @steinar: i guess this is a bug or at least very strange behaviour:
After investigating the internal avs file that the timecode node creates, it turns out that ffastrans in the "bad" case with milliseconds, ffastrans puts this line into the internally generated avs:
Code: Select all
m_clip = ShowTime(m_clip, 0, x=360, y=1040, font="arial", size=120, text_color=$ffffff)
and in the other case, which works:
Code: Select all
m_clip = ShowSMPTE(m_clip, fps=29.97, offset="00:00:00:00", x=360, y=1040, font="arial", size=120, text_color=$ffffff)
Unfortunately as i don't have the source code of ffastrans, i am not able to look for or guess what exactly is causing this issue, therefore i am not able to propose a workaround.
What i don't really understand is why you would use 2 encodings instead of only one... i guess you have some encode node somewhere later anyway, why not adding the -vf you use in the custom ffmpeg node there? Are you aware that when "omitting" the video and audio codec, ffmpeg will fallback and default to some very crappy low-quality settings, e.g. mp3 audio and low-bitrate mpeg4 video codec. So maybe you want to re-work your workflow anyway so you don't even run into this problem
cheers,
emcodem