alphaRonaldo
alphaRonaldo Compare new with old. Try watching them in headset, you'll see the difference.
yes, there is a difference i observed and loved the sharpness of those several old scenes
i can share with you more of the scenes where i noticed great sharpness of the image and fine details.
i was working on those scenes at that time:
https://www.sexlikereal.com/scenes/karaoke-night-28941 at the end of this scene you can notice out of focus part
https://www.sexlikereal.com/scenes/movie-night-28315
https://www.sexlikereal.com/scenes/big-black-tits-28255
I know why those scenes were sharper
Production at that time was setting aperture of the lens at 5.6
and the image is sharpest at that point
you can read about the effects here
and here
the tricky part here is that the focus at 5.6 is quite narrow and you have to be very carefull at every position and check the focus with every change of position
you can also notice that background is blurrier because it is not in focus
the issue with aperture 5.6 is that it is easy to miss the focus if you not very attentive and forget to change the focus
that happened several times already on several scenes, not only US scenes
i sadly cannot dictate the rules for production
so production decided to switch to the aperture 8
with aperture 8 you get a wider distance from the camera in focus and its harder to miss the focus
and as you can see there is a tradeoff to that
besides production part im not sure how many releases other studios have per week
but with previous and current quantity of the produced SLRO scenes we have to fit into release schedule
that requires us to think on the most proficient workflow that allows us to process enough scenes per week and at the same time preserve decent quality
there are many other nuances that regular users not aware of
such as codecs that we use to process the scenes
that we have 2-3 stages of the post processing
that there is a final transcoding on the server that limits bitrate to around 30 mbs
that some codecs preserve more quality but slower in processing
we researched that over the years and came to one universal solution
but it was already changed several times and we are again rethinking it
that all takes time and effort
together with responsibility of keeping the releases steady
of one thing i can mention for some thoughts is that "best" way to preserver quality of the video is to first denoise it and then apply color
to denoise it first and to keep the close to original color information of the video you have to encode to uncompressed formats such as prores
which just for the record would take around 30-35 hours per 20 minutes video on the best configuration computers that we currently have, latest gpu 4090, almost latest cpu's and motherboard and 60-120 gb ram
graphic cards are not involved as much when encoding to proress to it take that long
now if we split the scene to several computers lets say 3 that is still 30-35 hours for 1 hour scene
thats only one step
it goes after initial editing, qa, sound, correction and then denoise step
for the record one 1 hour scene files size is around 1 TB
you have to somehow copy that to those 3 computers or use local server that would provide sufficient bandwidth to allow you reading the files from 3 computers at the same time
the files produced after denoise to uncompressed formats is around 1GB per 1 second
thats 60gb per 1 minute
and 3.6TB per one hour scene
transferring such huge files between computers... ok, you got the point with the server i guess
After denoise there has to be either color correction or stereo correction or both at the same time
now if we make color correction separately it is another step and another uncompressed files which has to be accessible for the next step somehow
If we to make stereo correction and color correction that would increase the render time to about 1 day per scene
this step encodes the result into a compressed .mp4 format and it slightly reduces the quality as any encoding does if not to uncompressed
then dont forget another transcoding with happens on the slr server
i lost my thoughts train somewhere here
but there's a lot more i could explain on as many version we tested and their pro's and con's
but as a conclusion:
the current pipeline that we have is currently the best in terms of time/quality
there are some tradeoffs in production and post production
it is not like that just because we don't care
but it is like that because we came to that point through tests and current requirements
Still im again making one more research and hopefully it will give us some better results
