Original Effects Cleaned up with the help of AI...????

[Metryq]

But, at least presently, I think it would be better to simply recreate the original shots using more familiar 3D modelling and cgi animation, but set up meticulously to look exactly like the original aesthetic, but cleaner.

One thing some people may not realize is that Ai has no "sense" of 3D space. Hence, we get common Ai distortions like 6 fingers, additional limbs that warp and twist like rubber, or even intersect other rigid bodies. I see many Ai "renderings" where people merge with walls or cars, etc. like the final scene in The Philadelphia Experiment. In fact, common 3D packages can produce such intersections, where bodies pass through each other, unless the human artist avoids that, or is using "physics" simulations like collision detection. Ai is working purely from reference images, combining several that may that match. There is no "intelligence" in that Ai, no "understanding" of 3D space. I could go on listing other distortions I've seen, but you get the idea.

 

[Warped9]

One thing some people may not realize is that Ai has no "sense" of 3D space. Hence, we get common Ai distortions like 6 fingers, additional limbs that warp and twist like rubber, or even intersect other rigid bodies. I see many Ai "renderings" where people merge with walls or cars, etc. like the final scene in The Philadelphia Experiment. In fact, common 3D packages can produce such intersections, where bodies pass through each other, unless the human artist avoids that, or is using "physics" simulations like collision detection. Ai is working purely from reference images, combining several that may that match. There is no "intelligence" in that Ai, no "understanding" of 3D space. I could go on listing other distortions I've seen, but you get the idea.

Yes, I was talking about this very thing in my thread Unseen TOS in the Arts forum. AI doesn’t “see” anything. It certainly doesn’t see or understand three-dimensional space, weight, mass, gravity, time and the laws of physics and motion. It just perceives data as a two-dimensional collection of pixels without meaning or substance.

An object within a 3D modelling program and rendering engine is a construct of mathematically locked-in coordinates that are absolute. And the programming will not deviate or drift from those locked-in coordinates. But absolutes don’t exist for current generative AI. Nothing is truly locked-in no matter how specific and restrictive your prompts are. I tried to create a simple little video of my 3D model and it was an exercise in escalating frustration, even after providing orthographic views of my model and instructing the AI to use this as the sole reference source for the model. The instant it cannot “see” or “understand” something, no matter how seemingly insignificant, it defaults to its predictive behaviour and any similar reference materiel in its stored dataset and uses that as substitute—it starts hallucinating. And trying to correct the errors usually makes things progressively worse because AI has no real memory of what it just did a moment ago. It is essentially starting from scratch with each image you are asking it to generate.

Generative AI is presently a cheat or lazy shortcut for creating content as evidenced by the slop permeating Youtube and the internet. In many cases it isn’t up to the tasks being asked of it. It can be utilized with limited effectiveness in rather narrow ways. For example AI could not be an interior decorator, but an interior decorator could use AI to help a client see what their home could look like. A hair solon could use AI to show a customer what they could look like with a new hairstyle. A contractor could use AI to illustrate possible renovations to a customers home. In every case AI can be used as a faster and more sophisticated version of Photoshop, but it can’t imagine or suggest anything of its own—you still need the human professional for that.

The Large Language Model is proving to be a poor substitute for a real person who can actually reason, make value judgements and intuitively understands three-dimensional reality.

 

[Campe]

I’m not a huge fan of AI, but I am required to use it in my job. One thing I’ve learned is that it’s all how you write the prompt. Also, I would imagine that CGI artists would actually use professional grade software, not something readily available to the masses like ChatGPT or Gemini. I would think those tools would be more equipped to handle these requests. Perhaps there’s even a “keep the original intent” option. I would also hope they would involve people who have a history working with the franchise in order to get it right.

But, that’s all presuming this would even happen. It ain’t bloody likely if you ask me.

 

[Ssosmcin]

But, that’s all presuming this would even happen. It ain’t bloody likely if you ask me.

I expect no more work to be done on the original series at all. Once the 2006 work was fnished, that was it for the most part. Other than the few episodes barely tweaked for The Roddenberry Collection, the series was re-released at least 3 more times with exactly the same content. Just repackaged and re-pressed. Right down to the "coming soon" preview of the first Abrams movie.

The 2006 effects are more than good enough for people who can't see past 60's era work.

 

[Ragitsu]

I’m not a huge fan of AI, but I am required to use it in my job.

My condolences.

 

[Warped9]

I will admit that one day I would like to see what an episode of TOS could have looked like if they had had a bit more time and money. Given the resources of the day it wouldn’t look radically different, but it would have allowed them an extra layer of detail and polish consistent with what was then possible.

 

[Metryq]

Also, I would imagine that CGI artists would actually use professional grade software, not something readily available to the masses

Well, it depends on what you mean by "professional." CGI software does not have to cost thousands of dollars. Blender is an open source project easily the peer of professional packages. In fact, it is used by many professional studios, but that doesn't stop any hobbyist from downloading it, too. Of course, a professional studio like Pixar has software engineers on staff constantly pushing the boundaries, custom designing software as needed for each new project.

I will admit that one day I would like to see what an episode of TOS could have looked like if they had had a bit more time and money.

Ah, but limits of time and budget are just as much a part of the equation of "art" as ever. The real reason I get bent about what Lucas did to the original Star Wars trilogy—aside from many juvenile lapses in judgment—is that those movies are the way they were partly because of limitations of one sort or another. Necessity being the mother of invention, and all that. If Star Wars had been done in the inefficient manner of 2001: A Space Odyssey, it never would have made it to cinemas.

 

[Phaser Two]

Ah, the irony of applying AI to a show that welcomely dumped on human-usurping computers at every opportunity. Kirk can't pull the plug fast enough on such an idea.

 

[Warped9]

Current AI doesn’t need Jim Kirk to drive it crazy—it already has a built-in inclination for torrential streams of illogic.

 

[Metryq]

Current AI doesn’t need Jim Kirk to drive it crazy—it already has a built-in inclination for torrential streams of illogic.

Because it renders pretty flowers that smell bad.

 

[Metryq]

Gaussian splats may help one understand how generative Ai "creates" images. While the following videos feature Gaussian splats applied to 3D, remember that those scenes were captured in 3D—a sheaf of photos of a static scene "stitched" together like "panoramic" or "object" VRs, if you are familiar with them. Working purely from 2D images, even a sequence (motion) does not provide depth—unless the machine is programmed for that. And I suspect that will be next generation Ai.

In the following videos, Gaussian splats are a "point cloud" (or "splat cloud") that is, essentially, another form of data compression. You may have various senses for the environment around you, but perception is in the mind.

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

 

[Dar70]

Well it was just a thought guys. I was thinking of it as another tool in addition to other restoration/repair methods to make the image better and more coherent. Not make a totally new CGI image.

 

[Ssosmcin]

Well it was just a thought guys. I was thinking of it as another tool in addition to other restoration/repair methods to make the image better and more coherent. Not make a totally new CGI image.

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

 

[Warped9]

So I tried something out. Many years ago I ran a project thread in the Arts forum called Never Seen TOS Scenes where I photoshopped scenes of events in the TOS universe we never actually saw onscreen. One of those images was of Kirk and Pike together on the bridge when Kirk takes over from Pike. It came off well enough for the time, but just for fun I asked AI to clean up the image. Okay, to its credit it didn’t add anything or substitute its own version, but simply cleaned up and sharpened exactly what was in the image I had created.

I will say I was suitably impressed. It did a good enough job to make me think I could be looking at an actual screencap.

Small note: that original image I created has often turned up online (uncredited) used by some to depict Kirk and Pike meeting.

 

[feek61]

Well, where it is???

 

[Warped9]

Here you are. The top image is the original I photoshopped. The bottom image is the AI cleanup.

Part of what I find interesting about this is how it cleaned up Kirk. The image of Pike is a screen grab from the "The Cage." In that scene I believe Pike was talking to Spock. I swapped out Spock for an image of Kirk taken from "The Corbomite Maneuver" (if I recall correctly) and I had to try changing Kirk's tunic into the style and colour of that from the pilot episode era. The AI must have interpreted what I had done somewhat clumsily and fixed it.

I'm not going to bother posting it, but AI did generate one image that was...interesting. It put Kirk and Pike into the same shot together and all that was okay, but the bridge background was totally messed up. When I tried to get it corrected the hallucinations just got worse. If I were to do any more of these cleanups I would stick with the images I photoshopped myself originally and just get AI to clean them up.

 

[Warped9]

I was just thinking and being nitpicky here. In my image of Kirk and Pike together there are some incongruities. Firstly Pike would have to be about a decade or so older than how we saw him in “The Cage.” Kirk would look more like he did in “Where No Man Has Gone Before.”

Now I know little of naval protocol, but I wonder if transfer of command would actually happen on the bridge or would it be done more formally somewhere else. Maybe someone else could answer that.

 

[BK613]

Now I know little of naval protocol, but I wonder if transfer of command would actually happen on the bridge or would it be done more formally somewhere else. Maybe someone else could answer that.

When I was in, there was a change of command ceremony in front of the crew where the outgoing CO was formally relieved by the incoming CO. Before then, the crew answered to the former, and afterward the latter.
Doesn't preclude Pike showing Kirk around the ship as a courtesy prior to being officially relieved though.

(Oh, I look forward to your revamped image also showing up as a faux legit screenshot from years ago. )

 

[Warped9]

When I was in, there was a change of command ceremony in front of the crew where the outgoing CO was formally relieved by the incoming CO. Before then, the crew answered to the former, and afterward the latter.
Doesn't preclude Pike showing Kirk around the ship as a courtesy prior to being officially relieved though.

I looked it up. The transfer of command, at least in the U.S. Navy, has both formal and informal aspects to it. Informally the incoming and outgoing CO’s perform an inspection of ship and crew. The incoming CO is apprised of not only current status, but also current business or actions. Later there is a formal procedure where each CO reads out their respective orders in company of the assembled ship’s crew. I believe it finalizes with the incoming CO saying, “I relieve you.” and the outgoing CO replying, “I stand relieved.” A transfer of command can be spread out over a few days.

Now in TNG we saw a transfer of command when Picard is relieved by Jellico in “Chain Of Command” (if I recall correctly). What we saw looked like a rather pared down version of what seems to be standard in today’s U.S. Navy, but then again the transfer took place “in the field,” so to speak, with Jellico transferring over via a ship-to-ship rendezvous. If this had taken place at a starbase or back on Earth the transfer might have been a more protracted affair.

Or it’s just as plausible that in Starfleet a transfer of command has evolved into a more pared down procedure. Circumstance could also play into it if there happens to be an ongoing crises or situation at the time. There was certainly no ceremony when Kirk replaces Will Decker in TMP.

Finally we don’t know how the transfer of command from Pike to Kirk took place in TOS. It could have been back on Earth or at a starbase or even “in the field” with Kirk sent out to rendezvous with the Enterprise to take over from Pike. My head canon has Kirk in command of the Enterprise no more than maybe a few months in “Where No Man Has Gone Before.” Maybe Pike was expecting to take the Enterprise to the galaxy’s edge when he was promoted to Fleet Captain and command was transferred to Kirk.

(Oh, I look forward to your revamped image also showing up as a faux legit screenshot from years ago. )

That could well happen. The new image has a Gemini watermark, but watermarks can be easily removed.

 

[Sketcher]

Here you are. The top image is the original I photoshopped. The bottom image is the AI cleanup.

Part of what I find interesting about this is how it cleaned up Kirk. The image of Pike is a screen grab from the "The Cage." In that scene I believe Pike was talking to Spock. I swapped out Spock for an image of Kirk taken from "The Corbomite Maneuver" (if I recall correctly) and I had to try changing Kirk's tunic into the style and colour of that from the pilot episode era. The AI must have interpreted what I had done somewhat clumsily and fixed it.

I'm not going to bother posting it, but AI did generate one image that was...interesting. It put Kirk and Pike into the same shot together and all that was okay, but the bridge background was totally messed up. When I tried to get it corrected the hallucinations just got worse. If I were to do any more of these cleanups I would stick with the images I photoshopped myself originally and just get AI to clean them up.

In all honesty I much prefer the original image. Like most AI images, there is an unnaturalness that's jarring to look at. Yeah, the original is low-res, but it's clear there was work you and the rest of us should be proud of the effort that was put into.