Encoding timelapse footage

[Joachim Buambeki]

Hi,

I want to encode my timelapse footage with x264 at 2k and 4k in the future.
Since the clips are usually at max a few minutes, time is not really an issue.
I just used preset placebo with CRF 15 so far. But I wonder if there are settings that are useless for this kind of footage and can therefore be disabled to save some encoding time.
Flat skies are also more than common, so what Is prefered to avoid banding in them?

Best Regards
David

[poisondeathray]

I just used preset placebo with CRF 15 so far. But I wonder if there are settings that are useless for this kind of footage and can therefore be disabled to save some encoding time.


I would say "placebo" is kind of useless. It's called "placebo" for a reason :)

Look at your encode logs to see if a typical encode of this source actually uses 16 b-frames or 16 reference frames. If not , you're wasting a lot of time for nothing. You can get a general idea of what typical footage of this sort actually uses, and then customize your settings.

--slow-firstpass is kind of useless too. The usefulness of this have been discussed in several threads



Flat skies are also more than common, so what Is prefered to avoid banding in them?



Look into preprocessing with dithering filters. In avisynth there is one called gradfun2dbmod

[Blue_MiSfit]

Indeed. Be prepared to spend a lot of bits to keep the dither though :)

CRF15 should preserve nicely.

[mp3dom]

To preserve dithering (or to avoid it) you could set qpmin to a lower value (even 0 if necessary), disable fast-pskip and raise psy-rd (or keep it) in the range of 0.8-1.0. Raising psy-rd too much could create some sort of ringing. Anyway, like Blue_MiSfit said, prepare to use a lot of bitrate.

[Joachim Buambeki]

Thanks for your replies.
Look at your encode logs to see if a typical encode of this source actually uses 16 b-frames or 16 reference frames.
....
Look into preprocessing with dithering filters. In avisynth there is one called gradfun2dbmod
I had a look and there does not seem to be any benefit from going over ten b-frames at all. What line in my log tells me the number of reference frames used?
I already use Gradfun2DBmod through MCTemporalDenoise. :)

To preserve dithering (or to avoid it) you could set qpmin to a lower value (even 0 if necessary), disable fast-pskip and raise psy-rd (or keep it) in the range of 0.8-1.0. Raising psy-rd too much could create some sort of ringing.I will try that, thanks.

Best Regards
David

[JEEB]

Just as a mention, such noise as dithering usually does gain quality by raising psy-trellis in the --psy-rd option (the option was mentioned, but I remember some testing showing that more grain/noise/dithering was kept with the -trellis part of it) :3 Of course, it also might have negative effects on non-moving things, but if kept under control it should be completely fine.

[Joachim Buambeki]

Before I open a new thread for that, I figured this could be answered in this existing thread as well.

My timelapse footage is a sequence of sRGB TIFF frames. They are converted to YV12 at the beginning of my Avisynth script with "converttoyv12()".
Is that correct?

Do I have to specify rec.709 or something else like TV levels in my commandline to get a standalone BR player compatible output?

Best Regards
David