Upscaling videos

LOL, how did i miss this thread. This is exactly what I’ve been doing for the community for a bit now. There isn’t a magic bullet combination that applies to all videos. Combination of avisynth, AI upscaling, and AI interpolation is needed to get good results in a lot of cases. Search my post history for some examples. LMK if you have any questions.

Here’s a cost breakdown that I recently did on the back of a napkin. Costs are roughly for a 1-1.5 hour video put through AI upscaling and AI interpolation:

If a video takes me 10 hours of processing to do with 30 min of manual input, the math looks like the following.

  • 500 watt power draw X 10 hours = 5 kilowatt hours of energy

  • 1 kWH in my area costs $0.35

  • 5kWH * $0.35 = $1.75 in power

  • SF minimum wage = $16.32 per hour

  • 30 min @ $16.32 = $8.16

*** Power + Minimum Wage = $9.91**

2 Likes

Just curious what tools you use and general settings / advice for your process. Your stuff generally comes out very high quality.

I want to upscale some old 4k VR videos. Is it worth trying on VR videos?

Short answer is, try a little bit and see how it looks…

There are so many factors that affect if it is worth the effort. However, expect it to take a very long time to perform the upscale. It took almost 90 hours using a 16 core / 32 thread CPU @ 4.5GHz, and an nVidia 3080 Ti, to upscale and enhance a 1 hour 4K video to 6K when I did it. See details and some before and after pictures in his script release thread:
https://discuss.eroscripts.com/t/remaster-wankzvr-moka-mora-gimme-gimme-mora/47530

So 1 hour 4K → 6K took 90 hours. It will take considerably longer to render 8K. If you have less powerful hardware then you will add even more time to the work. Read the posts in the thread above where upscaling is discussed a bit more.

1 Like

worth is dependant on hardware and quality. for something that is already pretty high res, the worth is much more small compared to something that might be 480p or lower moving up to some HD resolution in both value and time.

If the VR video you had initially has a lot of problems (e.g not sharp, bad encoding) then its possibly worth it, just don’t expect it to be finished in a short time, as VR videos are inherently very large due to resolution and typically also long because of length, if you’re working on VR content that isn’t animated/CG

1 Like