Yes, that's quite similar to her look before I applied a PS filter, but her skin was a bit rougher (AI two years ago )

____________

Using a neural network, I was able to remove microphone noise when recording my voice for the second movie in the Lost Planet trilogy. This was a major problem for me 10 years ago. Today, I was able to remove the unwanted noise from the voice recording, and my movie now sounds much clearer.

The most important matter is always to find contentment in one's own work, as one's reward for the time spent creating. For example I prefer the look of the untouched lass on the left as well, but that's just my own nutty tastes, so no problemo. We all have different eccentricities. I'll add that the altered image on the right certainly has a style with commercial potential, which is a thing that should not go unappreciated.

Anyway, good luck on your forays into new technologies. I recall years ago watching the first part of your Lost Planet movie, and thought there was a lot of good work in there   , so I'll take a look at the latest installment when I get some time.

Hello to everyone who remembers me.

I'm trying to work with AI to create animated movies. I decided to imagine what the characters from my first movie, The Lost Planet trilogy, which I created in 2010, would look like if I were working on it today using neural networks.

To do this, I had to transform some individual frames from several scenes of this old movie into the artistic style of a realistic photo.



This is what the original frame from the movie looks like 20 years ago. The scene was created using bryce, Poser, and After Effects.



This frame was obtained using a neural network with the quality of a real photo.

As a result, I created the short video teaser from new scenes animated using a different neural network.

YouTube is completely blocked in my country today. So, I published my teaser on another similar video platform.

I really hope you won't have any problems opening my teaser. Here's the link:

Lost Planet (2010) AI Teaser (RuTube)

I want to understand: does it make sense to completely remake the first movie of the trilogy using modern neural networks?

Hi Poser - User,
Good to read from you in this crazy times.
I'd like to say, the two frames you posted show the problems of AI as well as the advantages.
The look of the AI image is great, almost "larger than life", but the couple don't even seem to communicate. The old shot, on the other hand, is full of life. Great eye contact, facial expression, body language. AI doesn't know much about human interaction, yet.
The trailer looks amazing, thou. The polished look works great for SciFi.
About a complete remake, I'm unsure, if it's worth the effort. I have the first two parts in my archives, and they are masterpieces, as they are. Especially the very first works great, in spite of, or just because of it's technical limitations. A remake would be very different, but I don't think, it would be better. Just interesting for comparison.
That are my thoughts about it. I hope, they are helpful.
Take care, my friend

____________

Thank you for your feedback!

I agree that artificial intelligence is still learning to understand human relationships. It doesn't even yet understand why we have a biological body, since AI doesn't have one, with all the ensuing consequences. But even the basic capabilities of AI allow us to create things that weren't possible just a few years ago.

For example, last year I completely corrected my recorded voice for my second movie. Ten years ago, when I was recording my voice for the film's characters with a microphone, I heard additional noise from the fans inside the computer. After editing and enhancing my voice in a digital editor, this noise increased noticeably. But at that time, I hadn't found an acceptable tool for removing noise from recorded voices.

Today, with the help of AI, I was able to correct both the voice and noise files. The voice came out completely noise-free, and I replaced the old files in the movie. The quality of the voice in the movie improved dramatically, and I published it on RuTube.

Lost Planet 2 (2016) RuTube

In this case, AI proved indispensable, and my movie significantly improved its sound quality.

In regard to the visual reworking, if you're pleased or intrigued by the AI filtering, then revising an older animation would certainly be a more efficient method of playing around with it, learning its strengths, weaknesses, and quirks.

If you're wanting honest opinions on the output comparison, my thought is similar to my earlier reaction. Both look like cg images, showing flavors typical of uncanny cg characteristics. Both images exhibit a general artistic style of realism, but neither looks much like a photograph of something in the real world. The particular visual style of both images is cg-realism, sometimes better called hyper-realism, which is different than photographic realism.

The original image had a quaint charm of uniqueness or even nostalgia, while the AI reworked image looks just like all the other AI imagery barrage flooding the internet. It's no slight against your own effort in this case; regardless of the input, this AI stuff all looks the same. My view is that makes it at best visually bland and uninteresting, but I appreciate that others don't share my view.

Going back to what I wrote a few months back, whether to continue in that direction depends on your own artistic and personal goals. If you're looking to make things that will appeal to current internet trends, then of course the AI look is very popular these days due to the work of the marketing people to promote it in search engines at the expense of everything else. There's nothing wrong with following current trends and styles-- most people in life are content to do little else. I think hobby artists should only follow trends if they genuinely prefer the style in question, as hobby is about passion rather than $$$, but that's not a common opinion, and not likely to get anyone upvotes on Reddit or other places of conformity and crowds.

Basically, my opinion is that AI filtering reduces the input of the artist, in favor of emphasizing 'the AI look' in the output. Certainly an AI image still exhibits choices of the artist in general composition, but in such visual matters as color saturation and lighting subtlety reduces the artist's input, changing everything to the preferred algorithm the AI is programmed to use. This is inherent in current approaches to AI, which is based on techniques of predictive consensus-- it's literally designed to process output to make it indistinguishable from everything else, with no originality or uniqueness.

On the other hand, it's very helpful that you were able to use AI audio filtering to accomplish otherwise impossible sound editing. That's the thing about new technologies-- we never realize their potentials until we start playing around with them, so the playing around part is an important part of the learning and comprehension process.