May Dev Blog
I went into details last month regarding how we deliver our game, versus others and how you the fans respond via supporting people like us. It's been made clear that the majority of you prefer idea complete updates, over deadline focused updates. This last cycle was a tough one and I will highlight a bunch of the problems it had.
Design, Design, Design
SR7 has been doing what he does solo for a few decades. With art there rarely is a team. When we ventured into games he remained sol. Then you add our deadline focus (Patreon billing) and planning was never expanded beyond a napkin. Ideas rapidly were implemented the day they were conceived, no prototyping, just implementing. By progressing in this fashion your mistakes are always behind you, but are always nagging you of their presence. For every situation you conceive bears the debt of completion in this system. Nothing becomes more evident of that than Themysicra. It was born prior to sweeping changes in direction and focus last July. It reeks of the of poor planning. However the debt has to be paid, and within that payment comes grind and burnout. Themyscira is not fun, it's a brutal environment, and were happy it's where it is now. But we must push on and finish what we started. So you all can expect more content for Themyscira in the near future.
Production as a role
It become extraordinarily evident that a production role was needed on our team. Implementing so many moving parts and trying to keep long term planning on track was impossible without oversight. The role really blossomed during this last cycle, and we are more confident than ever that we can achieve both idea completion and external deadlines. The team is now looking at every month as a potential release window, and that is very promising for everyone.
Investment In The Future
AI is a hot topic right now, for good of for bad. Some hate it, and think its a scourge, and others will toss $$ to people issuing prompts to a computer. We're left with no choice but to start and look at ways AI can help streamline our processes. We must preface that we will always be open in what tools we use in our processes, in fact using AI in general puts us in a precarious position. So with that in mind our two main objectives currently will be: alternative angled backgrounds, concept art, and prototyping. Let me explain below:
- For alternative angled backgrounds: It's difficult to plan time for these BG's. They come up during implementation/art finishing, take roughly 2-4hrs to complete (not including the idea phase). There is roughly 10 or per update. By having AI potentially generate something that can be touched up and used, we achieve massive time savings to spend on core art that matters. We will be creating an internal proprietary model for this.
- Concept art: If you play SU long enough you will eventually see the same pose from the same angle. If we can generate images that are close enough to SR7's style we can generate 1000s of images and pick several that have interesting angles. From there SR7 can use those images as references when creating scenes.
-Prototyping: We're actively creating a test bed right now to start testing out new mechanics and features. In this environment we do not want to steal time from art creation to create art for something you may never see. By using AI to generate concept art, it will help us rapidly prototype many things.
We're looking at AI as a tool, to be utilized in appropriate manners. In the coming months we're certain more and more creators will be utilizing AI in some form. It's our duty to you to acknowledge our usage prior to us implementing it, rather than be exposed later.
Comments (43)
Sorry for asking but when will the update to the main game cum out?
if, as you said, the artstyle is is close to your style, i think ist fine.
AI for backgrounds and references seems like s really good idea. Especially background taking so long and its not something you really pay that much attention to in game.
Setting aside how generative ai is ethically dubious. I think it sucks. It's creatively bankrupt in the worst ways. Every part of the process of making art is made with intention and creativity. Stripping them from the process to make things faster, for profits, is lame. Even if all it's used for is concept art, the cost is undeniable. It's like selling off an organ for money. I can't support this. As time passes and ai further seeps into the creation, I do not think I'll be renewing my patreon subscription nor will I be playing much more of this game.
I felt the same way initially. But then you go and consume your favorite media. Within that media you have technological advances that have benefited the creator in allowing them to express themselves further than previously. Digital art didn't strip the process, it allowed for the removal of the pen & paper, it expanded the tools, and helped creativity in unmeasurable ways. The invention of the word processor, increased efficiency, and allowed the internet to flourish. Flying with auto pilot didn't remove the need for skilled pilots. AI is a tool for us, not a means. We have the means and have proven so for a decade. Please do not compare us to those who don't have the means to create original art. If others can take our art and make it indistinguishable to ours (which is currently happening). Why would we be held to a higher standard, and not allowed to utilize that tool? Especially if we can expand on it from where it's at currently with the creative process. Lastly, I hope you stick around and see how we utilize it, before you pass more judgement. I think you will like how we will be utilizing it.
I mean im fine with you guys utilizing AI tools to make the workload better. My problem is I could immediately tell the korra pinup was AI, I assume that won't be the case for actual game assets.
That image is unedited and just prompted by an individual. It was used as an example of what we can edit. I can assure you that AI will eventually get to a place where unedited vs edited is indistinguishable its just a fact. But for us we will be using AI to edit, and the goal is for you not to know. The reason we announce our intentions is that it's better to acknowledge usage rather than get accused of it at a later date (which will happen to many creators).
If this means that a pose/backgorund from AI pics is just used for tracing or referencing, then I don't see a problem with this.
utilizing AI is fine, it's almost unavoidable for how beneficial it can be. Relying solely on it is crippling to a project
Exactly! Having an artist who's art is used to train the models is a tremendous advantage. It's almost like you get a worse version of SR7, and then SR7 comes in and fixes it up. Excited to see where this goes in the long term.
Man, AI is nuts. So you just fed it a bunch of your art for reference and it was able to pump out that picture? I'm curious how much direction it needed, or if it was just a "alright, now make something similar" and then BOOM - Korra pinup.
So that picture uses a model called pony xl then you use things called LORA's to help train the AI to mimic style. There are two publicly available LORAs currently, but other fan section individuals are doing their own as well. When prompting it really depends on the initial model, and how well the LORA has been trained. I ran a few tests copying prompts others shared and edited the prompt slightly. I could get some very decent images, however the consistency image to image is not there. The above image has been fine tuned using a few other inputs that do multiple passes over hands and or face. Image generation and prompting are going to be proprietary pretty soon.
Ow man, BoB2 will be a blast once its released.... just six more painful months...
AI is the future whether people like it or not. I am intrigued to see how you guys will use it.
I just saw a post where they can animate still images in SD. The technology is moving so fast, we can't even keep up and we feel we're early adopters.
Depending on the implementation of AI art, I may have to withdraw my subscription. We want you art, not some computer generated imitation.
That is fair, and we rather come out and acknowledge our usage and intentions early, than be accused of it at a later date. The later will happen as AI is getting very good at this point, and will not be worse than it is today. I do want to say that any image we generate and share will be using our own assets to make the model. The only difference being is the computer and artist now work together. The advent of digital art didn't ruin art it just changed how artists could express themselves.
If it gomes to stuff like (Adult Games) this i dont mind Ai art.
Honestly, I'm fine waiting longer for art assets if it's not done using AI, and I'm sure I'm not the only one.
We generally have the same sentiments, but in the environment now we have to take steps in order to keep up with eventualities. With how fast the technology is moving forward, we need to adapt and learn or we will fail. The assets being generated will be trained using 100% hand drawn images. In fact there most likely will be a time when actual skilled artists will draw for AI and then use the generations to help express themselves. I don't think this tech takes away from great artists, I think great artists have a massive advantage in utilizing the technology.
I noticed the game assets weren't divided into their usual segments for Themescera. Was that due to the rushed implementation?
In what version? The stand alone is implemented by Henshin, where as the Unity versions are SR7.
I’m definitely wary of AI, I do believe most of these models operate on stolen works from other creatives. If you’re using one trained on yours alone, and with care, I guess I can keep an open mind. But im Patreon because I’m interested in your art not a machines.
I don't want to make promises due to the evolution of the technology changing so fast. But I don't for see a point where we don't have a need for an artist. SR7 still has to come up with the art in which we train the models. Additionally his eyes, and knowledge are there to keep the art consistent and clean. But with that said AI in it's current state we're looking at just assisting. We believe in the near future, you're either going to acknowledge usage or get accused of usage.
Everything I was going to say or ask has been done. So I’ll just be excited and look forward to the updates and releases. Thanks for the info looking forward to the next update.
I'd have to say this is the correct use case for an AI assistant. Utilizing only the assets you already have, which are of a consistent style and known providence. Especially for background details, and as the basis for assets which are then touched up by the actual artists. I'm just hoping we don't dive into AI programming. Shouldn't need to here, but that's a domain where I remain unimpressed. Sure short scripts and the like can function, but past that not so much.
Not to mention utilizing an in house model for R34 removes some of the murky legal issues. Especially where they intersect with crowd funding, fan art, and porn.
AI might help to hopefully give more animated scenes by providing the in-betweens. Looking forward to the future.
Wait Soo Themyscira would be released in June!?
Just curious which dedicated AI system did you decide to invest in?
We bought a PC with a 7950x, 128mb Ram, 4090, plus dedicated UPS to run training over night when necessary. For art currently we're exploring the use of SDXL
Just be careful. AI is fun to play with, and I’d argue that porn is one of the best uses for it, but it’s a slippery slope, and we all signed on for YOUR work.
Yes and as stated these models are trained using the over 6000+ images we have in house currently. Using the AI to assist in production is far different than relying on AI for the production in our opinion. One uses an artist and the other doesn't.
Excited for whats to come
I really like it ❤️
Loved
Will you be investigating the use of AI for code development, compiling, etc?
We haven't yet. But I'm sure it's something we will explore the more complex we go with prototyping.
I am tentatively excited the AI stuff can lead to great things or slacking work I hope it’s the first love your guys work hate to see if faultier to AI. 👍
The thing is you need art to train the models. Even if we use AI to generate all the images, we would need to create the art to train the model. However we're only looking at AI assist at this point.
As good as the AI art is, the original has a magic to it
We agree and that is why we're currently only interested in AI assist. Plus when you have such a talented artist as we have, why not utilize that with AI.
The fact that’s AI generated, is beyond crazy. Good work lads