I would like some general advice on Video Upscaling for a beginner

The program is very easy to use. But something tells me there is depth and complexity almost like an art form. I don’t want to Upscale a video from a hamburger to a cheeseburger. I would like some advice on how to upscale to the best possible 60 fps most detailed video. I have been reading posts and I glean a little here and there but honestly its a bit overwhelming.

So in a nutshell as if you were explaining to a simpleton what general advice would you give to Upscale most videos? Example: I have a 360p 150 MB Video. I want it to look like a Bacon Royale with Cheese and I am willing to rerun the same file as many times as necessary to probably achieve this. I imagine it being in the ballpark of 3-9 times. What Preset Enhancement Programs would you run on the video file and in what general order would be ideal? I am hoping practice will make perfect but tips would be most welcome. If I missed a obvious FAQ (Other than the Youtube tutorials please do forgive me).

I think the main thing is to manage expectations. People see the buzzword “Ai” an immediately think it can perform miracles.

Garbage in - Garbage out.

You cannot take a 360p video and make an insane, high detail, crystal clear glorious 4k video from it.

1 Like

That is a very excellent response, thank you! I get it, no miracle working here. But in general what Preset Enhancement Programs would you run on it, in what order. And what would a reasonable expectation be? I have never really been a fan of 4k. Would a nice 1080p be reasonable?

Again that’s the fickle thing with video encoding, it’s literally different for every video. They’re like snowflakes no two are the same so there really isn’t a one size fits all.

If your 360 clip is super clean then it’s possible to maybe get a passable 1080 from it, sure. But in my experience asking for more than double the initial resolution tends to not come out so great…again many factors to determine, your millage may vary depending. Only sure fire way is to play around & find the limits for your situation.

If it’s a long clip i’ll cut out a 10 min chuck and run a bunch of tests to see what works best. You can run different models on different scenes but that requires you to split video, run your different models on said scenes then rejoin it again after. I’ve also found better results (for me) by running a ridiculous crf of like 7 or even zero, then cutting it down to size with Staxrip or your other preferred utility.

1 Like

I have found that VEAI has several short-comings. I listed them out here.
I think the one thing I don’t mention in there is that to increase frame rate, the best model is Chronos - SloMo. To get 60 fps from a standard 24~fps source, you have to run it at 300% into an image folder. Then convert that back to a movie with something like ffmpeg. (I can give you a sample command/batch file script to get you started if you want to use this approach.)
It’s a shame really. The Chronos Fast model is suppose to streamline the process, but it adds huge blurry patches over trees and other things quite often.

Also, in that post, I mention how mp4 format loses quality at any CRF. That is true for my computer monitor, but I can’t see the artifacts on my TV screen. I have a nice monitor, but it’s far from being accurate enough for professional movie editing and such. My TV on the other hand, is the cheapest thing money can buy. Anyway, there’s a chance you won’t even be able to notice the artifacts.

I would like to use your approach because I noticed several classic videos with 60 fps are so crisp and realistic. Please do share. I convert it with chronos at 300% to an image then use FFMpeg (Is that another Program choice?) to convert it back to a MP4?

Also at the bottom of Options a option gives you a scale of 0-30. It says something vague about the lower the more detail. Can you explain to me exactly what this is?

Yes. ffmpeg is a command line program.
Usually you can find compiled versions of it quite easily online.
You’ll want to do two passes with VEAI: Upscale and clean, then SloMo. (If you’re source is clean enough, doing SloMo first will save a lot of time.)
Once that’s done, the command for ffmpeg will go something like this:

"path to ffmpeg.exe" -framerate 59.94 -i "path to image folder\%6d.jpg" -map 0:v -map 0:a? -map 0:s? -map 0:d? -map 0:t? -r 59.94 -c:v libx265 -crf 20 -preset slow -c:a copy -c:s copy "output path\output file.mkv"

I’m on Windows, so this is how it looks for me.
-framerate should be the source framerate times 3.
-i is the input path
-map tells it what to do with audio video and things like subtitles. ? has it ignore if there is none. (Mainly this it just leftover from converting video to video.)
-r will set the output frame rate. It will drop or add duplicate frames if it’s not the same as -framerate. (For getting 60 FPS, it will drop a lot of frames, but I think the output still looks smoother than the original.)
-c:v is the video codec. I’m using H265 here.
-crf and - preset are settings for H265.
-c:a and -c:s are left over from video to video conversion.
output.mkv puts it in an mkv container. You can change that to mp4 if you want. It’s just easier to add the audio back in with mkv. Heh, there’s probably a way to pull the audio from the original file in the same command, but I haven’t looked for how to do that.

Anyway, all this can be done on Mac too or in bat files or shell scripts. You just have to change the command accordingly.

I am going to try this! Yes I use WIndows. Thank you

You convert film done in 24 fps to 60 fps ???

To be honest, only because that’s what my TV is limited to. I would do more if it did more.
Most of the movies I have are ~24 FPS stuffed onto a DVD at ~30 FPS. It’s not like these programs being used don’t know how to time the frames correctly when going up or down in frame rate.

i’m trying to understand, because it’s sound weird to me. you talk about cinema theater movie, which have been filmed in 24fps, that maybe some of them already maybe got a remasterisation or a FHD release, and you convert them in 60 fps ?

So far, they don’t make remastered versions of movies in 1080p 60 FPS. When they do make 4K remasters at 60 FPS, they’re using tools like VEAI anyway. They just have a lot more tools and skill.
24 FPS with motion blur has always been less than ideal. Either the blur is so much that the details get lost, or the motion gaps are so wide the result is noticeably jerky or flickery. For most people, it’s an unquestioned tradition. Same with film grain.
Anyway, yes. I like to convert things to 60 FPS when I can. I would do more if my TV could handle it. Sure, it’s not as good as if it had been filmed in 60+ FPS, but it’s better than 24.

Most people watched movies on TV for decades with the common PAL 50 hz interlaced fake doubled fps, so when anyone suddenly watches a movie in its original low fps he might feel it looks quite choppy.

Which Program on VEIA do you prefer for the clean up movie phase before the Chronos?

Also is it better to Upscale 200% and THEN clean up, or Clean up and then Increase Size?

It’s in fact the reverse which happened. You rediscover the movie, because everything was wrong before (except at theater), it’s as much different than moving from dubbed movie to original voice. on travelling moment, it’s difficult to ignore the difference. Having cinema at home is an amazing / weird experience. (Video projector).

Lately, Proteus. You have to turn off Auto-detection model parameters in preferences. Anyway, I usually look for a scene where there are a lot of farther away things that should be in focus. I hit the Auto-detect button then adjust from there. The dehalo slider always goes up with auto-detect, but is rarely needed.

Really it’s up to you and what you’re dealing with.

Having a 360p source will limit the greatness of your results by a lot.
The Gaia models are the least artifact heavy. Sometimes I like them for that.
Artemis medium is perfect for cartoons, I think.
There is no one model that fits all.
If you are super serious about it, you’ll divide the movie into scenes and find the best model for each scene. Most people don’t want to take that kind of time though, so you end up taking losses here and there.

Oh, so in your life most people went to the cinema not watching TV in the first place? :eyes:

in my place and everybody on the planet except in russia, china, india, and some other place, in “my life”, people are watching a movie at first in cinema theater BEFORE TV…
Movies was coming to Tv after 3 years,in the 80’s/90’s, then, after 6 months on DVD in the 2000’s.
it’s only recently that this time has been reduced to 4 months with VOD and SVOD.

if you watch a movie on tv before its theater release, then you know what you did, and so this conversation end.

in the 70’s/80’s the qualiity of movies displayed on tv was so bad (black and white, no stereo or 5.1), analog etc that’s everything was anecdoctic. the issue of framerate conversion was anecdotic compared to the other problem. this started to change with Digital received / cable / satellite.
I remember watching Terminator 2 on an official PAL VHS, and next to it i had the NTSC transfer of the tape bought in U.S.A. the transfer was so bad on the Secam/Pal version that we almost couldn’t watch the movie, and back then, we were thinking that this was an acceptable quality.
I had to see the quality of NTSC transfer to understand how bad we were.
quality movie / picture and understanding it for normal audience is very recent.
what we have today is amazing, and being able to enjoy such a quality today is a gift.

Sorry Didier, I am not speaking of the vilolation of legal law to “watch movies on tv first”. As said, the vast majority of movies is watched on tv in 50 hz interlaced with the fake double fps. Cinema always was too expensive to visit often and the tv at home is near and easy to access.