Using AI to generate graphics/music/etc for Dink
I was wondering how people feel about this? Absolute yea or nay, something in between?
Personally, from an ideological viewpoint, I must say I well and truly don't give a shit. If the end result is good it's fine by me.
But that's kinda the rub, you need to put in effort to get good results with today's AI. (Maybe in the future this changes, who knows.) If you just ask some online image generator for a picture and paste it in, it's not going to be great. A lot of adjusting, fine-tuning, and curating is required. At that point, it doesn't feel substantially different to me from using some other tool that can automate things, like texturing with a graphics program. The only type of computer art where you're truly not "cheating" in any way is fully hand-drawn pixel art.
Personally, from an ideological viewpoint, I must say I well and truly don't give a shit. If the end result is good it's fine by me.
But that's kinda the rub, you need to put in effort to get good results with today's AI. (Maybe in the future this changes, who knows.) If you just ask some online image generator for a picture and paste it in, it's not going to be great. A lot of adjusting, fine-tuning, and curating is required. At that point, it doesn't feel substantially different to me from using some other tool that can automate things, like texturing with a graphics program. The only type of computer art where you're truly not "cheating" in any way is fully hand-drawn pixel art.
I think it would be quite novel for there to be a d-mod generated almost entirely with AI tools, just to see if it's feasible. The engine is simple enough that several well-written prompts with an LLM could be used to generate maps and scripts, along with some sort of ComfyUI workflow for tiles and sprites.
Can't say I'm a fan, but that's mostly because the sort of people who use generative AI to make things generally seem to not actually care about doing the post-generation stuff, like curating, testing, and tweaking. Using generative AI as a starting point doesn't seem bad, but trusting the end result to be good on its own has routinely lead to disappointing products.
Music seems like the easiest thing to get generative AI involved in, as it's fairly good at doing that already and Midi seems like a more forgiving file-type than most. Graphics would be a challenge (at least if your goal was Dink-style graphics), and script writing would likely be a nightmare (for anything longer than a short romp).
Still, if someone wants to have a go, I wouldn't want them to cease on my regard. Even if they end up making something subpar, it's not like most of us haven't made something like that all by ourselves. And if they find a way to make something good out of it, all the better.
Music seems like the easiest thing to get generative AI involved in, as it's fairly good at doing that already and Midi seems like a more forgiving file-type than most. Graphics would be a challenge (at least if your goal was Dink-style graphics), and script writing would likely be a nightmare (for anything longer than a short romp).
Still, if someone wants to have a go, I wouldn't want them to cease on my regard. Even if they end up making something subpar, it's not like most of us haven't made something like that all by ourselves. And if they find a way to make something good out of it, all the better.
I think the most difficult part of making graphics for Dink is to make something that fits its art style and graphic quality, and doesn't stick out immediately as alien to the game. If AI can be used to get there somehow then I'm fine with it. I've never tried it, but judging from what I've seen in Necromancer I'm not impressed, to be honest.
I'm ok with using it for music too if that helps people to find something more fit for the tone of what's happening on screen, instead of just throwing in a MIDI version of Grenade by Bruno Mars as the theme of their main village.
What I've been using it for is story related things, because that's where I struggle with.
Not asking AI to write the story (it is not good at that), but help with general research, writing tips or some touch-ups, like "can you turn this next sentence into 1600's pirate speak?"
I'm ok with using it for music too if that helps people to find something more fit for the tone of what's happening on screen, instead of just throwing in a MIDI version of Grenade by Bruno Mars as the theme of their main village.
What I've been using it for is story related things, because that's where I struggle with.
Not asking AI to write the story (it is not good at that), but help with general research, writing tips or some touch-ups, like "can you turn this next sentence into 1600's pirate speak?"
I dislike AI generated stuff. Maybe because I'm someone who could be called an AI-sceptic. The more likely reasons are:
I've seen many people using AI because they're too lazy to create something original themselves. Though I have to add, those people were mostly con artists and scammers. I guess this is more of a problem of people abusing AI, rather then using it. I can understand why a regular person would use AI. Maybe they lack talent, they're disabled, or they might just have something missing most other people have, and are literally unable to create a particular thing. I have no issue with that, as long as they don't try to take credit for something unaltered that an AI created.
My other issue is that AI generated content just lacks certain things that only a human artist is capable of producing. I simply think AI image generation today isn' at at the level to compete with true human art. Sure, it can do many things, but it has it's limitations. This is why I think it currently can only threaten a smaller slice of artistic jobs.
Well, feel free to leave your own thoughts.
I've seen many people using AI because they're too lazy to create something original themselves. Though I have to add, those people were mostly con artists and scammers. I guess this is more of a problem of people abusing AI, rather then using it. I can understand why a regular person would use AI. Maybe they lack talent, they're disabled, or they might just have something missing most other people have, and are literally unable to create a particular thing. I have no issue with that, as long as they don't try to take credit for something unaltered that an AI created.
My other issue is that AI generated content just lacks certain things that only a human artist is capable of producing. I simply think AI image generation today isn' at at the level to compete with true human art. Sure, it can do many things, but it has it's limitations. This is why I think it currently can only threaten a smaller slice of artistic jobs.
Well, feel free to leave your own thoughts.
I agree with RangerLord 100%.
As someone who has used AI to make a DMOD (Necromancer), well, I'm for it I guess.
It couldn't do nearly as much as I wanted, but it helped as a springboard for things. Like the save machine. I used AI to generate a Mayan-esque circular thing, tried getting it into a 3D model, but ended up creating the 3D model almost from scratch using a Gimp edited version of the AI image. The blacksmith character was generated using MeshyAI and then imported into Blender to animate (and fix various texture error issues), a few other props with generated using MeshyAI, but in the end I'd try and find 3D models already made and free then modify or edit those.
I gave up trying to get AI to write DinkC scripts.
I did use AI to bounce around ideas and things, and then take the things that resonated the most with me and go off on my own tangent.
Some of the music tracks were generated using Suno AI.
Some artwork (like the backgrounds for the Magic/Items inventory screen and the level up screens were edited versions of AI generated iterations. As was the choice selection sprites.
And the Decretius disappearing video in the opening was AI generated, but not the Dink smashing the basilisk scale (that was all blender models etc, animated by me).
I assume AI's ability has increased now it's been about 4-6 months since I last mucked around with it, and that to get good results you probably have to subscribe/pay some company for access to the version that can make consistent characters etc. Hmmm, maybe it can even do walk cycles now....
It couldn't do nearly as much as I wanted, but it helped as a springboard for things. Like the save machine. I used AI to generate a Mayan-esque circular thing, tried getting it into a 3D model, but ended up creating the 3D model almost from scratch using a Gimp edited version of the AI image. The blacksmith character was generated using MeshyAI and then imported into Blender to animate (and fix various texture error issues), a few other props with generated using MeshyAI, but in the end I'd try and find 3D models already made and free then modify or edit those.
I gave up trying to get AI to write DinkC scripts.
I did use AI to bounce around ideas and things, and then take the things that resonated the most with me and go off on my own tangent.
Some of the music tracks were generated using Suno AI.
Some artwork (like the backgrounds for the Magic/Items inventory screen and the level up screens were edited versions of AI generated iterations. As was the choice selection sprites.
And the Decretius disappearing video in the opening was AI generated, but not the Dink smashing the basilisk scale (that was all blender models etc, animated by me).
I assume AI's ability has increased now it's been about 4-6 months since I last mucked around with it, and that to get good results you probably have to subscribe/pay some company for access to the version that can make consistent characters etc. Hmmm, maybe it can even do walk cycles now....
I'm firmly in the anti-AI camp. The ecological impact of it is my biggest grievance. Also, I really dislike how the economy is betting everything on its success- as if it's inevitable.
If it's about the usual arguments against AI of art theft via training data, or job loss due to AI replacement then I would say it doesn't really matter for Dink related projects. Though, I think it's still problematic.
I just can't personally justify the amount of energy and water AI uses for such lackluster results. I'm not sure I could even justify the cost for good results.
There are a few exceptions for where I find AI use acceptable or even encourage it's growth(one such place being the field of medicine) but overall I find it to be pretty awful and will be completely disinterested in any project that uses it.
If it's about the usual arguments against AI of art theft via training data, or job loss due to AI replacement then I would say it doesn't really matter for Dink related projects. Though, I think it's still problematic.
I just can't personally justify the amount of energy and water AI uses for such lackluster results. I'm not sure I could even justify the cost for good results.
There are a few exceptions for where I find AI use acceptable or even encourage it's growth(one such place being the field of medicine) but overall I find it to be pretty awful and will be completely disinterested in any project that uses it.
