It upsets me a bit, that I've spent so much time working on the consistency of the character, but the moment I post some unpolished stuff, it seems to be forgotten.
I feel that some people maybe have missed me saying that those new scenes are just sketches. If you want consistency and polished look — I've been posting updates for quite some time now, and in one of my previous posts I've showed the difference between WIP and finished work (it was in relation to facial expressions).
I'd say Hermione on the bottom looks rather different from the base design. Different proportions, different hair color... and too photo-like instead of a CG-model. Not saying it's bad or anything, but I feel like consistency is important. If it wasn't - AI-generated games would be much more popular than they are )
Well, yes, it's based on a different DAZ model entirely, I just repurposed an old client work, I liked the pose there and didn't want to do it from scratch. Not to mention that it was done in 10-20 min.
I feel like the issue with AI image generation is that people think they will be able to create photorealistic images, when 99% of the time they just look weird, compared to DAZ where you always know its a render and thats fine (besides obviously most of the games where people just render trash on their potato PCs).
I really like your artwork and hope you'll continue this awesome game, but the image in your post is a prime example of this. Looking at it, while it looks great and way better than what most people are able to generate, you just feel that something is off but you just cannot figure out what it is (the uncanny valley).
What bothers me the most and I think is where a lot of the backlash is coming from is that everyone feels the need to use AI. I work in IT and every single tool now has to have some kind of use of LLMs, with 99% of them just being trash. Things like code generation, describing things or helping with error messages and the results are almost never helpful (unless you are doing very very basic stuff)
Again, image in the post was intended to show a workflow, and not to present a finished piece.
The facial expressions are the thing that is currently finished (well, I'm integrating them into the main character file, but the expressions are done), and in my personal view they are as consistent as it gets (granted there's always room for improvement, but still).
As for LLMs, I think our experiences might be different here. I remember running GPT-2 locally a year or so ago. While I was fascinated, it was more of a curiosity than a tool. Now I'm genuinely using both local models and ChatGPT-3.5 to help me with the code, scripts and dialogue.
Granted, I'm an artist who had to learn how to code, and not a coder who writes code for clients, so I can't talk to that.
However, Chat-GPT was very helpful for me learning Python and a bit of CSS, while the code isn't perfect (technology is still very new), but I find it to be so much faster than googling all of the errors or trying to find a solution on stack overflow.
These are the examples of the CSS style I did for TextGenUI.
And this is an example of the same UI with the default style:
I was so exited to be able to do that on my own!
Would somebody who've been using CSS for let's say 5 years be able to do the same thing better and faster? I sure hope so, but I definitely couldn't do that, and with AI I was able to write something that I needed and now use regularly.
I've also used AI to help me writing some web-crawler + resize app to create datasets for LoRA training. While I ended up not using it eventually (there are so many LoRAs available now already) I thoroughly enjoyed developing it and again, it is highly unlikely that I would've even attempted doing that without AI.
I feel that the same approach I use for image generation could be useful for writing code (or any other text based activity really).
AI isn't too good at complex tasks or sequencing them, so I break the process down into smaller stages (just work on the eyes, just work on the mouth, etc.), the same could be said about writing code, while it would struggle with the whole app, it can handle separate functions.
I realize that this is the subject that people are having strong opinions about, but I can't help but to be exited, and AI is definitely an interesting topic to discuss.
Anyway, I appreciate everybody who shows interest in my game and takes time to voice their opinions.
