Home Creators Posts Import Register Favorites Logout

Content

So I've been playing around with GenAI for making an animated prelude to Silver Wings. Wrote the script a few years ago when there was some interest in doing a show/film based on it, but it never went anywhere... so now I'm going to use it to learn how the GenAI systems function and whether they're worth bothering with on this level.

This isn't a new full time project, my writing will remain focused on Bastion for the summer or until it's done. But I figure I can make a 20-25 minute episode over the run of a couple weeks, working an hour or two a day, based on the animation I just posted.

My Question for this post is... would you prefer a more realistic look, or a clearly animated look?

Both have advantages and disadvantages but neither have much of an edge over the other in terms of approachability. So it's purely what would you guys think vibes best with the source material.

For example, a young Sorilla Aida, aboard the Los Angeles :

The realistic option tends to lean softer than I'd prefer for Sorilla, I expect because the images it was trained on were mostly models, actresses, etc... Still good, still a good look for Sorilla that hits all the right points... but the vibe may be a little different than you'd get in the books. Which is par for the course when it comes to film casting I suppose. I would not complain if Silver Wings got a TV or Movie and this was the casting, but I would note the differences.

Now, the animated version, same character concept, same basic prompting.

Here you can see we get a harder look, it feels more muscular, more defined. I'm not certain that it's exactly Sorilla... but honestly I think that's more cause I tend to leave the characters kinda nebulous in my own head so that I have room to make adjustments if I come up with a cool concept later.

I like both concepts, I like both animated and realistic styles, personally. I think I'd prefer the realistic style if the systems were more consistent... then I thought maybe animated, but honestly that look MIGHT make the inconsistencies stand out even more somehow. Not sure.

What do you guys think?

Files

Comments

Patrick (Doc) Doherty

I like the realistic version best. Maybe it’s that the animated version looks pe-od. I also think that both need to be a tad more muscular maybe 15 -20lbs more muscle

Evan Currie

the issue is that the systems aren't yet ready for that level of fine tuning... well, Google's system actually IS... and it's really impressive... but it's not as good at the overall LOOK as Midjourney... and Google will flat out refuse to make changes like "Make the character a little more muscular" when working on photos you upload, this is clearly an attempt to prevent people from using their system to make AI deep fakes. GenAI systems are impressive, but the more I use them the more I'm honestly BAFFLED that big corporations adopted them so quickly. They're really useful for lone creators, but a big company should never use them. It's cheaper and better to pay someone to do the job if you have that option.

Hugh Blair

I much prefer realistic

Mooserman

I prefer realistic

Simon Chambers

I don’t think this will be biggest issue, it will be getting people to see it. Not sure if it is same person or multiple different people creating channels on YouTube but due to ai YouTube seems flooded by ai stuff, would not mind if it was original but i see the exact or closely similar versions of same story 100 times. It has got to the point I block whole channels

Adrian Burchell

i prefer realistic, if anyone has seen Noctari, something like that. seeing stuff put on youtube, maybe the huge film companies may be a thing of the past

Joe

I prefer realistic

Richard Bateman

Try Kling. Start with the free tier and see what you think. There's also Stable Diffusion. It has a free tier for strictly non commercial use. If you like it, you'd probably have to go to the $20/month to use the images the way you do.

Brion

More realistic for sure. But it is likely that she would be more muscular based on the things she does. Can you not prompt the AI engine to add more muscle?

Evan Currie

I'm not intending on some huge audience for this, nor am I planning an AI channel or the like. AI is a tool to extend my capabilities. Nothing more, nothing less. While I do try and "write" a story with AI, roughly once a year, that's purely to see the state of the tech for writing, not to replace my workflow. I write better than the AI, simple as that. So I will never use AI to write. It would be silly of me to give up my personal style for, as you say, a product that's just a repeat of a hundred other people. The script I have for this is one I wrote, originally because there was some interest in doing a Silver Wings... animation, I think it was? That fell through, but the script is still here. So I'm working on it with AI... mostly so I can learn the workflow, and see how it turns out. The stories will always be my originals.

Evan Currie

Not really, annoyingly. I've tried a bunch, but I don't think most of them were trained on muscled female models. I could be getting something wrong though.

Evan Currie

Not quite yet, but they're going to have to COMPETE again... and this is where AI really comes into its own, and we'll see real artists get serious work too... because how will the studios compete? not with AI, cause anyone can use that. How do you stand out? You go practical.

Evan Currie

I'll definitely go SD eventually, just cause I want a local server for as much as I can get. Kling is a bit better than what I use, but also more expensive. I'd love to use SeeDance 2, but that's pricier again. If this works out, I'll see if I can justify the... 3-6k per "feature" that SeeDance would cost. But first, I'm going to keep working with Omni cause it's good bang for the buck and I'm figuring out my workflow for the toolset.

Simon Chambers

Yes that is the way ia should be used. If you are going for ai go realistic. It just bugs me the misuse and the effect it will have long term on genuine writers starting out

Evan Currie

Maybe, but only if AI improves quite a bit more. Right now, for writing, it's... genuinely meh to bad. Now, if it gets a fair bit better, then yeah... we're going to risk losing some skills, and that would suck... but it's not the first time it's happened. We lose skills all the time when new tools render the old skillsets obsolete. In fact, now, we're NOT going to lose skills entirely, if only because tons of people take the time to record themselves doing the various skills. So in the event that someone wants or needs to use them, most skills today are recoverable. This is a historical oddity. But yeah, it's possible that in a couple more generations, people will "direct" novels instead of writing them... and they'll automatically be released as books, audiobooks, and "motion books" all at the same time. And THAT will definitely mess with how the next couple generations tell stories.

One4TheRoad

I myself prefer realistic. With advances in AI I can't see any reason you couldn't make the full movie. It would take a while mind you but it would be a great film

Evan Currie

Not quite there yet, I'd say, for a few reasons. Mostly the size of generated clips make it hard to maintain a cohesive and coherent flow. You'll see when I post the video I'm working on... i"m up to two and half minutes, and it's not bad, but every 4 to 10 seconds there's shift. Sometimes extremely subtle, easy to miss, sometimes... not. Now, some of this can be solved with practice and familiarity with the tech... but not all.

One4TheRoad

Logan Paul made this one https://youtu.be/wNKaYvhTauM?is=-KqLumE0u64VetE9 I myself just took 10 photos from a written book and asked AI I think Gemini to make it into 10 sec clips. With no prompting it was pretty good. I imagine with time and patience and very detailed prompts you could make something excellent. Just upload the pdf and go.

Nallus

I'd go with realistic, though a lot larger files I assume. Also, I would think Sorilla would not be pretty, but average, also muscular, and with a no-nonsense look 95% of the time.

Evan Currie

sizes are the same regardless, though POSSIBLY you could compress the animated one more after the fact. As to looks, probably right, but that's why I called it "Hollywood Casting". GenAI has an easier time with good looking models, likely cause that's what most of the training material was.

PaquiXochitl

I’ll be honest, I don’t like the animated version at all. When reading the books, she had a young Rosario Dawson vibe to me. But she’s too old for the part these days. I’m not saying animated wouldn’t work, I just don’t particularly like this one. Yes, as you observed, the realistic one is a little softer than one would imagine, but can’t that be fixed? I’ll be interested in hearing what your other readers think. I hope you follow up in another email. I would really enjoy seeing this series brought to film, but I imagine the expense would be enormous. I’m still a book behind on the series because I always get Kindle and Audible and go back and forth. Unfortunately, there never was an Audible version recorded of book ten. Will it ever be done on Audible? I’m guessing no, since it’s been a few years. Please let me know. I will just go ahead and read the Kindle since I already have it. B. A. Kidd-Hoffmann