Advice on How to improve as a User of AI Tech?

So First I will list what I do, then I will ask for advice.

  1. I color correct, voice correct

> 2) I put the video in the latest codec H265 Based on advice will move this step to after 3.

  1. Then I typically use Atremis, up the file to FHD, up the FPS to 60.
  2. Usually a file will go from 1G to 10G+ at this point so I hand break it at 22 Quality
  3. Which usually results in a 1GB file far superior to the original.


Do you see a Issue with my current order of operations?

Should I use Proteus, Thea or Gaia After Step 3 but Before Step 4?

Is it ok to use all 4 on a single video and if so what order would you use them for maximum impact on end video quality? Probably Artemis → Proteus → Gaia → Thea

With Thea does it make a difference if I do Fidelity or Details first?

What if I used all 3 would that result in the highest possible quality Version of the Video?

Should I consider 120 FPS? 60 Seems so Superior to the starting point, I have used 120 but 60 seems reliable and faster. I am not sure if the juice is worth the squeeze.

I have a Video with bad Interlacing and the Option helps but doesnt fix it all, about 20% is left, any tips? I was thinking run it again?

I have a Video with a hazy effect from bad filming it almost looks like Fog, I want to remove that and make the video Crisp, what would you do?

I would love any tips and advice you could give Thank you.

Step 2 should be avoided at all costs! Probably you are re.compressing an already compress source and removing even more details and sharpness. If you don’t have storage problems use at least ProRes HQ after step 1. File sizes will be bigger but the quality of the original source will be there and using ProRes is faster than using H265.
I will not use Artemis, maybe Proteus or Gaia HQ and use only 1 of the models to upscale in 1 go to your desired resolution.
Blending/Merge 2 models together is something I do a lot.
If you like the 60Fps result stay at 60Fps… less frames to encode in the end


Ahh Thank you so much, what you say makes sense. I will look into ProRes HQ.

As for not using Artemis, the types of Videos I am Upscaling have a LOT of issues. Mainly because they are from 20ish years ago lol. Proteus and Gaia just do not come close to cutting it on them. Their impact is much better on higher quality videos. I mainly deal in low quality noisy ones.

I will be honest the thing I am most curious about is the Thea Fidelity and Details. At what point would you run it? Before Artemis? After? Would you even use them and if so would you mainly just use details?

I’ve tried to get Theia to work for me, and it always leaves vertical lines. It might be better with a FHD source, so maybe you could try a pass after getting up to that resolution.

For at least one old show I have, I really liked the result of Artemis High Quality. BUT, I have to use the command line to get it to select the right resolution of model. Basically, the GUI generates commands that force it to choose a bigger model than needed for DVDs to FHD. (Like if you look at the benchmarks, they list out the speeds of 1X, 2X and 4X; it’s picking the 4X where it should be picking 2X and the difference is extreme.)

In general though, the Artemis models add these fake circle-spots on things like gravel, dirt, trees and grass.

To get around that, I use some other tool to clean the noise up as much as I can without cleaning out all the details. Then I fine tune Proteus manual settings and I can get some really good results.

I like Apollo for interpolating, but it is slower than Chronos (Not Fast. Don’t use Chronos Fast.), and I would do it after upscaling because the quality is that much better—though it probably takes more than double the time to process.

1 Like