[LilTKit / LittleTgKitten] Discussion of his work

[popexxx]

I just saw on Patreon that it's called NewBee and they're still developing it I think

Get more from NewBee on Patreon

Get more from NewBee on Patreon. Become a member for just €5/month. Doing TSF stories for the fun of it and expanding my creativity with my audience…

www.patreon.com

 

[chalana56]

I just saw on Patreon that it's called NewBee and they're still developing it I think

Get more from NewBee on Patreon

Get more from NewBee on Patreon. Become a member for just €5/month. Doing TSF stories for the fun of it and expanding my creativity with my audience…

www.patreon.com

Yep story is still progressing. Her patreon handle is newbeetsf

 

[boringname]

I don't understand what you mean at all. At first, she was a strong woman with a broad face. And then she became a thin, fragile teenager with a very, very narrow face. Yes, some of the changes could have been attributed to feminization, and no one would have noticed. But this is like the Melissa situation, where they slipped us a completely different model between shots. Therefore, I think it's inappropriate to refer to the loss of the original model when there was no attempt to recreate it.

But here's what I believe: the author may have genuinely disliked the original model. And the way she looks now was the original goal. It's just that at some point, it became clear that it was impossible to gracefully transform her, and the leap occurred.

Yeah, it's a terrible analogy. The correct analogy would be: you lost the essay on the computer, but you had a draft of it printed out on paper. So it would not be that hard to recreate the paper, just type what is on paper back into the computer.
In this case, she still has the rendered outcomes of the model. She could (with admittedly a little work) rebuilt the model to resemble the output that she has. She choose not to and opted for a different look in the new model. That's an artistic choice, not a "it's impossible to recreate the model" situation.

 

[viscountsheep]

Yeah, it's a terrible analogy. The correct analogy would be: you lost the essay on the computer, but you had a draft of it printed out on paper. So it would not be that hard to recreate the paper, just type what is on paper back into the computer.
In this case, she still has the rendered outcomes of the model. She could (with admitted a little work) rebuilt the model to resemble the output that she has. She choose not to and opted for a different look in the new model. That's an artistic choice, not a "it's impossible to recreate the model" situation.

just treat it like the time they replaced henry cavil with chris hemsworth's brother for geralt of rivia.
it's difficult to recreate a model even if you have an image because daz has too many morph keys.

 

[lexa]

Hey everyone,

I’ve been watching this discussion again with one laughing and one crying eye. And honestly, I keep thinking: if the problem really is that our main character changes too much visually from panel to panel, there is actually a pretty practical way to improve this.

LoRA — Low-Rank Adaptation. Before we spend another five pages arguing about her face , maybe this is worth looking at.
The nice thing is that experimenting with this doesn’t necessarily require some huge expensive setup. There are open-source tools like Draw Things and others that can be used for this, and if your own hardware isn’t powerful enough, GPU time can also be rented online.

So why do I think a LoRA could help here?

Think of it as teaching the AI: “This is HER.” Instead of trying to describe the same woman again and again with prompts, you train a small model on carefully selected images of that particular character. Face, hair, proportions, characteristic features, general appearance. It doesn’t magically guarantee 100% identical results in every possible situation — AI is still AI — but it can make character consistency much, much better.

And the interesting part is: The material is already there. We now have so many panels and renders of our heroine that there should be enough good reference material to build a carefully selected dataset. Pick the strongest and most consistent images, prepare them properly, train the LoRA… and use that as the reference going forward. And because most of the comic already seems to be rendered anyway, I wouldn’t even suggest replacing the existing workflow. Quite the opposite.

3D render → Img2Img → character LoRA → final panel

The artist could still create the scene, pose, lighting, clothes, composition etc. exactly as before. Then the existing render becomes the Img2Img reference, while the LoRA helps pull the character back toward the established appearance. That seems much more useful to me than expecting AI to invent the whole panel from scratch. Would it need some testing? Of course. Would the dataset and settings matter? Absolutely. Would it solve every single consistency problem with one click? Nope. But could it be a practical way to make our heroine look much more like the same person from panel to panel? I think so.

And this is the part I sometimes don’t understand about discussions here: We can spend pages saying “her face changed again” or “she doesn’t look like the same character anymore.” Okay. Fair criticism. But then let’s also talk about possible solutions.

Constructive criticism + an actual idea for fixing the problem.

That moves things forward. Everything else eventually becomes a little bit… bla bla bla.
Just an idea. Maybe it’s useful, maybe it isn’t — but I thought it was worth putting on the table.

XO,
Lexa

 

[viscountsheep]

Hey everyone,

I’ve been watching this discussion again with one laughing and one crying eye. And honestly, I keep thinking: if the problem really is that our main character changes too much visually from panel to panel, there is actually a pretty practical way to improve this.

LoRA — Low-Rank Adaptation. Before we spend another five pages arguing about her face , maybe this is worth looking at.
The nice thing is that experimenting with this doesn’t necessarily require some huge expensive setup. There are open-source tools like Draw Things and others that can be used for this, and if your own hardware isn’t powerful enough, GPU time can also be rented online.

So why do I think a LoRA could help here?

Think of it as teaching the AI: “This is HER.” Instead of trying to describe the same woman again and again with prompts, you train a small model on carefully selected images of that particular character. Face, hair, proportions, characteristic features, general appearance. It doesn’t magically guarantee 100% identical results in every possible situation — AI is still AI — but it can make character consistency much, much better.

And the interesting part is: The material is already there. We now have so many panels and renders of our heroine that there should be enough good reference material to build a carefully selected dataset. Pick the strongest and most consistent images, prepare them properly, train the LoRA… and use that as the reference going forward. And because most of the comic already seems to be rendered anyway, I wouldn’t even suggest replacing the existing workflow. Quite the opposite.

3D render → Img2Img → character LoRA → final panel

The artist could still create the scene, pose, lighting, clothes, composition etc. exactly as before. Then the existing render becomes the Img2Img reference, while the LoRA helps pull the character back toward the established appearance. That seems much more useful to me than expecting AI to invent the whole panel from scratch. Would it need some testing? Of course. Would the dataset and settings matter? Absolutely. Would it solve every single consistency problem with one click? Nope. But could it be a practical way to make our heroine look much more like the same person from panel to panel? I think so.

And this is the part I sometimes don’t understand about discussions here: We can spend pages saying “her face changed again” or “she doesn’t look like the same character anymore.” Okay. Fair criticism. But then let’s also talk about possible solutions.

Constructive criticism + an actual idea for fixing the problem.

That moves things forward. Everything else eventually becomes a little bit… bla bla bla.
Just an idea. Maybe it’s useful, maybe it isn’t — but I thought it was worth putting on the table.

XO,
Lexa

what?

it's a 3D rendered comic. the neat part is that you have explicit control over every aspect of the character's appearance with the character model.
a character model by definition retains it's likeness unless the user changes it. so it's identity and appearance are always retained.
preferably you'd backup the models of the character you'll be editing frequently, if you don't and your working model gets corrupted or "lost" we end up in a situation like this.

it could also be that the male base model's topology fighting the feminine shape morphs but daz has a universal body mesh from gen 9 so it's unlikely that'd be the issue.

if you're gonna use AI and LoRAs, there is no need for a 3D render in the pipeline.. you can just use full AI pipeline at this point. this would double or triple the effort needed for a panel with model config, image masking, multiple iterations to get the desired look, maybe even a depth controlnet or preserving the poses,
all for better resemblance ? at this point the 3D render would be just a guidance data with less than 50% weight. nobody with a sane mind would be willing to do that.

i'd rather do a face transfer with the daz plugin from the older render to the newer model and be done with it. then again it's the author's choice, if they simply feel like the new model is more apt for the narrative then we just gotta trust the process.

 

[lexa]

what?

it's a 3D rendered comic. the neat part is that you have explicit control over every aspect of the character's appearance with the character model.
a character model by definition retains it's likeness unless the user changes it. so it's identity and appearance are always retained.
preferably you'd backup the models of the character you'll be editing frequently, if you don't and your working model gets corrupted or "lost" we end up in a situation like this.

it could also be that the male base model's topology fighting the feminine shape morphs but daz has a universal body mesh from gen 9 so it's unlikely that'd be the issue.

if you're gonna use AI and LoRAs, there is no need for a 3D render in the pipeline.. you can just use full AI pipeline at this point. this would double or triple the effort needed for a panel with model config, image masking, multiple iterations to get the desired look, maybe even a depth controlnet or preserving the poses,
all for better resemblance ? at this point the 3D render would be just a guidance data with less than 50% weight. nobody with a sane mind would be willing to do that.

i'd rather do a face transfer with the daz plugin from the older render to the newer model and be done with it. then again it's the author's choice, if they simply feel like the new model is more apt for the narrative then we just gotta trust the process.

Thank you!

And actually… THIS is exactly what I was talking about.

A constructive reply. An explanation of why an idea might not be the best solution, followed by another approach that could work better.

I genuinely appreciate you taking the time to explain it.

My LoRA idea may have been unnecessarily complicated for a 3D/Daz workflow — fair enough! And your point about simply doing a face transfer from an older render to the newer model sounds like a much more straightforward solution.

I learned something from your reply. And that’s exactly how a discussion should work, in my opinion.

Someone throws an idea into the room, someone with more experience says “Actually, there’s an easier way…” and suddenly everybody reading the thread knows a little more than before.

I also hope that one of the creators occasionally reads discussions like this. Maybe there’s something useful in them that can be taken away — whether it’s my original idea, your correction, or somebody else’s suggestion.

That seems much more helpful than creators having to read through pages of bashing about what they supposedly did wrong.

Just saying…

And yes, ultimately you’re absolutely right about one more thing:

It’s the author’s choice.

If the change is intentional and the creator feels the newer model fits the direction of the story better, then we simply have to trust the process and see where it goes.

Thanks again for the constructive response. Much appreciated!

XO,
Lexa

 

[ganzer]

Visual artists are in an AI crisis, this new survey explains why: L.A. arts and culture this weekend

A new survey shows AI is deeply impacting the careers of working artists, a majority of whom feel they are in direct competition with the technology.

www.latimes.com

Visual artists do not like AI, unsurprisingly. The technology is destroying their careers and their industry.

99% of respondents out of a pool of 378 verified professional visual artists noted their dislike for the technology, with 92% categorizing it as “strong dislike.”

Even more jarring, the survey found that 80% of respondents believe that they’re in competition with the technology. The artists expressed deep concerns about the impact AI is having on their careers, with 54% saying it has diminished their income, 75% their job and clientele security, and 90% their income opportunities. Another 77% said it had negatively impacted their career growth, 61% the future of their career, and 74% their career sustainability.

No offense, but statistically were you to actually tell a 3-D graphic artist "You should just use AI for your modeling" the overwhelming odds are they would consider it both tone-deaf and insulting.

Not commenting on the value of the idea, what is happening in the industry with the proliferation of generative AI tools is regrettable but also unavoidable. Whether or not artists like them or not they are going to be part of the game going forward.

 

[Communist Bimbo]

The difference between an artist and commercial consumer is that artists actually enjoy the process of creating art. The creation of art is about the process and about the meaning of the art, not the desire to simply put out soulless product. AI exists because the wealthy want to make you dependent on only them for your 'entertainment' needs. They want to steal from you the ability to create art, both on a skill, time and financial-level. If we have the time to create our own art, that is bad for the wealthy. If we have the time to create our own art, that is bad for the wealthy. If we have the financial freedom to focus on our art, rather than dedicating all of our time to working the jobs that they decide are morally and financially correct for us to be working, that is bad for the wealthy.

They hate you, they have me, and they will always hate us. AI is just their latest way of attacking humans, our expression, our culture and our ability to exist without them controlling our lives and thoughts. It's disgusting filth.

 

[viscountsheep]

i don't agree with the "AI bad" narrative.

established artists hate it because it threatens their income stream. now every moron who never held a pencil can make images without commissioning people who draw as a career.
i think it lowers the barriers in art.

a couple of decades ago if you needed a good chisel or ceramic plate you'd have to commission a blacksmith or a potter, hopefully they accept the job, pay them what they ask for, be patient till they make the thing, and be happy with whatever you got. it's a process spanning a week or two

now you can just walk into target and pick up from a collection of stuff in standardized sizes and shapes. pick from a greater variety without the long wait. sure it's not hand made, it might not have a soul and human touch to it. you might not get it exactly how you want it.
now the people who used to make it for higher price has to lower their rates to sell stuff. they can't churn out similar looking medium quality products at mass like the industries do. so they either find a better job or get good at what they do.
that being said there are professional toolmakers and potters who charge a fortune for custom made stuff. and it is still a valid line of work.

it's just that now the industrial revolution is happening to digital art.

i can absolutely hand draw a panel for a comic in 4-6 hours(cause obviously im bad at it) a 3d render on my potato pc with a 1650 at 0.9k with iray takes 25 minutes per image(and the laptop is basically frozen for that period)
OR i could generate a 1K image in 3 minutes and do some manual touch up in the next 20 minutes for a full panel.

i genuinely think AI democratizes art.

you just need to move with the world. focus on making a DIY cupboard after picking up the nails and planks from the store instead of trying to cut the tree, saw and craft the planks, forge the nails and finally putting everything together to make a hand crafter cupboard. yes a handcrafted one will have more soul to it. but for an average joe it's more practical to pickup materials and tools from the store.

this is a hill i'll willingly die on

 

[Nuvolari]

I think AI can be a tool for artists. I have even seen some genuine art that used AI tools. But it does not really "democratizes art," because in 99.9% of the cases what people put out does not rate as art. It does not even democratizes commercial illustration. It democratizes crap.
Let's look at one example from these forums. Joe Six-Pack, who HAS some experience in doing artwork, has been dabbling in AI comics. Now, you might argue that Joe is not a great artist, and I would grant you that. But I have seen the results of his AI dabbling, and I much prefer his non-AI work. It may not be great, but it does its job far better than the soulless, cookie-cutter stuff that the AI spews out.
I recommend people to read Scott McCloud's seminal work, "Understanding Comics: The Invisible Art", in particular Chapter 7: "The Six Steps." AI is fairly good at Step 6, mediocre at Step 5, crap at steps 3 and 4, and utterly hopeless at steps 1 and 2. Yet those earlier steps are the most important of all.

 

[viscountsheep]

democratization of art means more people can now create visual media, i think nitpicking the semantics of what art means does nothing to prove otherwise. to me a failed art project is still an ART project.
your phone camera democratized photography. i don't think anyone would hire a crackhead with a smartphone to do their wedding photos.

Ai is a tool. a tool is only as valuable as the person using it.

the hammer doesn't need to apologize for not building a crate when a chimp picks it up and starts banging it against a pile of wood imitating a carpenter.

great recommendation btw.

 

[lexa]

Well, this turned into a much bigger discussion than I expected.

But I actually enjoy reading the different opinions here. And maybe this is the more interesting question behind all of it. Where does art actually begin?
With the hand that holds the pencil? With the technical skill? With the idea? Or somewhere in the whole process between an idea and the finished work?
I understand the argument that AI is threatening jobs. And I also understand why an artist who spent years learning how to draw might not be very excited when somebody comes along and says: “Look, I made this in 20 minutes.” That is a real issue. I don’t want to dismiss it. But I’m still not convinced that this automatically makes the result worthless, or the person behind it completely non-creative. Let’s say I have an image in my head. I develop the idea, describe the characters, decide on the composition, atmosphere, lighting, clothes, expressions and what I want the image to communicate. Then I spend an hour prompting, changing things, rejecting results, correcting details and pushing it closer and closer to what I originally imagined.

Did I paint it?

No.

But did I have a creative role in creating it?

I think I did.

Maybe that’s where the definition is shifting. A photographer doesn’t manufacture the camera. A tattoo artist doesn’t build the tattoo machine. A digital artist doesn’t write the rendering software. And a 3D artist doesn’t necessarily model every asset from a cube before using it.

Tools have always changed. Photography was once attacked as something that would destroy painting. Digital art wasn’t always taken seriously. 3D rendering certainly had people saying “the computer did it.” Graffiti spent decades being treated simply as vandalism and today some of it hangs in galleries.
None of those comparisons are perfect. AI brings completely new questions with it, especially around training data, copyright and people’s jobs. I’m not pretending those problems don’t exist. But human creativity has always developed together with its tools. So maybe AI is another shift in where some of the work happens. And this is where I struggle with putting everything into two boxes: “real art” and “AI crap.” There is terrible hand-drawn art. There is terrible photography. There are terrible 3D renders. And God knows there is an enormous amount of terrible AI output. But surely there can also be good work made with all of those tools? For me, the interesting question isn’t simply:

“Did AI make this?”

It’s:

“Was there a human idea behind it, and did someone use the tools available to turn that idea into something worth looking at, reading, feeling or thinking about?”

Maybe the person doing that becomes less of a traditional craftsman and more of a director, curator or designer. Fine. We can argue about what word to use.
But I find it difficult to say there can be no creativity involved simply because the machine performed part of the execution. And maybe viscountsheep’s hammer analogy gets close to something important here. A powerful tool in the hands of someone without an idea can produce rubbish very efficiently. It doesn’t follow that the same tool in the hands of someone with a great idea must produce rubbish too. Anyway, my original LoRA suggestion clearly wasn’t the smartest technical solution for this particular 3D workflow. I learned that from the replies. But I have to say, I quite like where the discussion went from there.

This is much more interesting than another five pages arguing about Dee’s face.

Lexa