00:00
00:00
Newgrounds Background Image Theme

CapacitorCat just joined the crew!

We need you on the team, too.

Support Newgrounds and get tons of perks for just $2.99!

Create a Free Account and then..

Become a Supporter!

Is anybody actually using AI in their workflows now?

1,128 Views | 51 Replies
New Topic Respond to this Topic

it depends on what im doing at any given time. early on i tried seeing if i can use it for rough drafts of a particular design i'm looking for, but i find that 90% of the time, i have far more control over it when i stitch something together myself. that isn't to say that there aren't niche cases in which i would try it out and see what happens. it's not an overly large time investment just to see if it potentially spits out something you can work with.


its probably best suited for writing & research related things more than anything i would ever upload to this website


signature

BBS Signature

if i was retarded, i would definitely use it


collect my pages

BBS Signature

Nah, not for me.

I kinda enjoy doing the scraping of other people's work myself, tearing that shiz up and putting it back together weirdly myself as well, then squeezing out that steaming hot stinky slop, using my own farticacial insmelligence.


But who am I kidding here? I dun got no workflo, yo!


Short answer, no. Long answer, no, and I will not touch the mediocrity engines until the day I die


he/they - pretty cool guy - yt channel here - Flash Controller Mapping Crew


My nuanced take on the whole generative AI used in filmmaking debate is that its done for one of several reasons:

1) They are control freaks and don't like the collaborative process. I met a guy in college who did his entire thesis with AI simply because he thought everyone else was "too ego starved" and would try to change his idea too much.

2) It's faster and requires less effort and creativity, like you said. Corporate initiatives have deadlines, and AI removes the need for brainstorming or coming up with ideas, so its faster overall. Additionally, AI is trained to do things in the most palatable way possible: so it's great at avoiding alienating people.

3) People see it as a viable option for cost-cutting and consistency. An AI can't revolt or anything like that, so having it pump out a dozen cartoons in the time it takes the average person to do a minute of animation is good, never mind that you don't need to pay nearly as much as a salaried employee. I get the argument that, for ametures, it can help bridge the gap between skill and vision, but I've always found that its better to learn how to do something well by starting out really shitty at it.


It'd be stupid to dismiss everyone using GenAI as "dumb techbros". Coca Cola is using it for their advertising now, and I figure in like 5-10 years it'll be sort of in a similar situation as the Paul brothers are seen now: yes, most rational people find them to be moronic douchebags, but some brands still use them to advertise their stuff and there are a surprising amount of people I've seen online who defend them vehemently. AI is like that to me, in that most people will still hate it in the future, but there are companies who will continue to use it.


Me personally, I'd rather see it used for more blue collar-type jobs: like having it explain how to fix things around the house yourself rather than paying someone to do it for you, diagnosing mechanical problems with your car to save the mechanics time diagnosing problems (or so you can do it yourself), and maybe have it curate a personalized list of news stories for you in the morning. In such a way, it just streamlines normal work processes rather than replacing jobs entirely.


BBS Signature

At 11/27/25 04:33 AM, Skoops wrote:I've been asked to implement it in my workflow at my job, I learned as much as I could stomach since I was being paid to do so. Ultimately, as someone that was already competent at my job, it didn't make anything easier or faster, it just crammed a slot machine in the middle of my workflow, a major timesink to get something different than I imagined. It was always the bottleneck, not the "tool" people keep insisting it must be. After my trial period, I told my bosses what I already knew: it's not worth it. Can't make anything that would be made public because everyone would clock it as AI and it would damage the brand, can't use it for internals with clients because it makes promises in preproduction that we can't keep in production. Better to have sketchy and not-so-polished designs that we understand vs something that looks perfectly rendered at a distance but makes no fuckin sense up close.


That must've been a pain in the ass trying to work around that shit. I imagine it's like working with a co-worker that you have to fix all the fuck ups for, but can't exactly teach to not make the same mistakes. Businesses going all in on that is going to backfire. Hell it already has when Valve just recently caught Blackrock's lawyers using AI in a legal battle. The stupid machine hallucinated points and evidence that didn't exist.


Not to mention, the final product cant be copyrighted or profited from without a possible lawsuit.


At 11/27/25 02:47 PM, ProsciuttoMan wrote:It'd be stupid to dismiss everyone using GenAI as "dumb techbros".


This is true. Even talking with some of the people who are currently in the graphics design industry have been put in a position where they had to learn it.


Whether or not people feel one way or the other about it is another story, but I'd imagine that a lot of people are either in the same boat as I am where, in spite of all of the flaws Ai has choose to take a neutral stance of some kind OR point out that it's not really doing anything to improve their workflow as Skoops described in his post.


Coca Cola is using it for their advertising now, and I figure in like 5-10 years it'll be sort of in a similar situation as the Paul brothers are seen now: yes, most rational people find them to be moronic douchebags, but some brands still use them to advertise their stuff and there are a surprising amount of people I've seen online who defend them vehemently.


This is why I've brought up practicality as the key issue to focus on when it comes to Ai and workflow.


Sure Coca-Cola is using it for their advertising but that doesn't mean it's producing good results. Actually this is why in one of the last threads on Ai I brought up a news source on how some companies have ended up hiring the artists they fired to clean up the glaring issues in the results Generative Ai makes. <- This in of itself should be a sign that it's not all cracked up as a lot of people are hyping it up to be.


AI is like that to me, in that most people will still hate it in the future, but there are companies who will continue to use it.


Some may, but in my honest opinion I think it's only going to remain a behind the scenes type of thing.


I do not anticipate that it will take over the industry like a lot of people are fear mongering about for that reason. Personally I wish that people understood the angle I'm taking from a business perspective but that's not exactly easy when most people don't have that understanding regarding production, either.



At 11/27/25 03:21 PM, Dolorious wrote:
At 11/27/25 04:33 AM, Skoops wrote:I've been asked to implement it in my workflow at my job, I learned as much as I could stomach since I was being paid to do so. Ultimately, as someone that was already competent at my job, it didn't make anything easier or faster, it just crammed a slot machine in the middle of my workflow, a major timesink to get something different than I imagined. It was always the bottleneck, not the "tool" people keep insisting it must be. After my trial period, I told my bosses what I already knew: it's not worth it. Can't make anything that would be made public because everyone would clock it as AI and it would damage the brand, can't use it for internals with clients because it makes promises in preproduction that we can't keep in production. Better to have sketchy and not-so-polished designs that we understand vs something that looks perfectly rendered at a distance but makes no fuckin sense up close.

That must've been a pain in the ass trying to work around that shit. I imagine it's like working with a co-worker that you have to fix all the fuck ups for, but can't exactly teach to not make the same mistakes. Businesses going all in on that is going to backfire. Hell it already has when Valve just recently caught Blackrock's lawyers using AI in a legal battle. The stupid machine hallucinated points and evidence that didn't exist.


It's something like that, yeah. It's like having an employee that you have to train from the ground up every time there's a new task, only he doesn't understand your language.


I used ComfyUI for image generation, which is supposed to be the most fine-tuneable shell for Stable Diffusion, and I guess it is. It requires heaps of plugins and tangled webs of node-based instructions to have any chance of getting what you "want," especially if you need one thing to look a particular way and another thing to look a different particular way in the same image. You can't just bash different generations together because they change to a different random angle and perspective every time unless you're doing a "controlnet" which is a whole can of worms in and of itself. Suffice to say, you'd need to be most of the way done with doing something by hand if you wanted it to "help."


So it takes a few hours to get set up and another 30+ minutes of pulling a slot machine handle before you get something that's close enough. You'd think that once you're all set up with the node layout and everything that you wouldn't have to do it again, but nope, every time you gotta get a new type of image, 20% of that setup needs to be changed. There's never a point where it gets easier. The knowledge ceiling is fairly low, but the fact that it sometimes randomly does what you tell it to do (despite all the reasons it can't do that consistently) makes people think that the knowledge base is deep and you can wrestle its limitations into submission with enough custom nodes and whatever. It's amazing how superstitious and ritualistic the people that actually like this shit are; they've all got their own little magic tricks and folk remedies that they swear by that don't actually change the odds at all.


All that to generate a base for what I'd then have to work on manually.


So, like, at this point I know more about how to use it than anyone at my company despite being the one that hates it on principle more than anyone. My bosses still think it's a magic free labor machine, and I do what I can to dispel that delusion. They only ever see the successes on the internet and assume people are getting those on the first try, so I keep all the failed generations and time logs saved so I can show them how many failures it takes and how much time it wastes trying to get something close enough which, again, does us no favors in actual production. Of course, I can't get through to them all the way. They think whatever hard limitations it has now will just magically go away someday soon, but as someone that's spent enough time with it, I'm pretty sure those things are baked in. It feels a lot less scary when you actually experience just how much you have to fight against it to get anything that isn't broken, let alone what you asked for.


At 11/27/25 10:58 PM, Skoops wrote:
At 11/27/25 03:21 PM, Dolorious wrote:
At 11/27/25 04:33 AM, Skoops wrote:I've been asked to implement it in my workflow at my job, I learned as much as I could stomach since I was being paid to do so. Ultimately, as someone that was already competent at my job, it didn't make anything easier or faster, it just crammed a slot machine in the middle of my workflow, a major timesink to get something different than I imagined. It was always the bottleneck, not the "tool" people keep insisting it must be. After my trial period, I told my bosses what I already knew: it's not worth it. Can't make anything that would be made public because everyone would clock it as AI and it would damage the brand, can't use it for internals with clients because it makes promises in preproduction that we can't keep in production. Better to have sketchy and not-so-polished designs that we understand vs something that looks perfectly rendered at a distance but makes no fuckin sense up close.

That must've been a pain in the ass trying to work around that shit. I imagine it's like working with a co-worker that you have to fix all the fuck ups for, but can't exactly teach to not make the same mistakes. Businesses going all in on that is going to backfire. Hell it already has when Valve just recently caught Blackrock's lawyers using AI in a legal battle. The stupid machine hallucinated points and evidence that didn't exist.

It's something like that, yeah. It's like having an employee that you have to train from the ground up every time there's a new task, only he doesn't understand your language.

I used ComfyUI for image generation, which is supposed to be the most fine-tuneable shell for Stable Diffusion, and I guess it is. It requires heaps of plugins and tangled webs of node-based instructions to have any chance of getting what you "want," especially if you need one thing to look a particular way and another thing to look a different particular way in the same image. You can't just bash different generations together because they change to a different random angle and perspective every time unless you're doing a "controlnet" which is a whole can of worms in and of itself. Suffice to say, you'd need to be most of the way done with doing something by hand if you wanted it to "help."

So it takes a few hours to get set up and another 30+ minutes of pulling a slot machine handle before you get something that's close enough. You'd think that once you're all set up with the node layout and everything that you wouldn't have to do it again, but nope, every time you gotta get a new type of image, 20% of that setup needs to be changed. There's never a point where it gets easier. The knowledge ceiling is fairly low, but the fact that it sometimes randomly does what you tell it to do (despite all the reasons it can't do that consistently) makes people think that the knowledge base is deep and you can wrestle its limitations into submission with enough custom nodes and whatever. It's amazing how superstitious and ritualistic the people that actually like this shit are; they've all got their own little magic tricks and folk remedies that they swear by that don't actually change the odds at all.

All that to generate a base for what I'd then have to work on manually.

So, like, at this point I know more about how to use it than anyone at my company despite being the one that hates it on principle more than anyone. My bosses still think it's a magic free labor machine, and I do what I can to dispel that delusion. They only ever see the successes on the internet and assume people are getting those on the first try, so I keep all the failed generations and time logs saved so I can show them how many failures it takes and how much time it wastes trying to get something close enough which, again, does us no favors in actual production. Of course, I can't get through to them all the way. They think whatever hard limitations it has now will just magically go away someday soon, but as someone that's spent enough time with it, I'm pretty sure those things are baked in. It feels a lot less scary when you actually experience just how much you have to fight against it to get anything that isn't broken, let alone what you asked for.


Yeah sounds about right. Heard very identical stories from programmers/coders who were told to use AI to generate code for whatever software from their boss. Having to go through the slot machine process, then spend many hours fixing all the fuck ups the LLM generated. Even coders are annoyed with this shit.

But yeah LLM's are schizo as you can tell with it's randomness and it's concerning that doctors, lawyers and other important professions are starting to rely on it.


Heard a recent story of some guy using AI to try to identify edible mushrooms. Man uses the machine to examine a mushroom. Machine says the mushroom is safe. Man eats the mushroom. Gets fucking poisoned. I doubt we will have to worry about anything like Skynet. The closest we're going to get is the Cyberpunk 2077 AI's. The ones that act like full schizo viruses. Except it would lean more towards envisioned Y2K style of fuckery.


At 11/28/25 12:10 AM, Dolorious wrote:
At 11/27/25 10:58 PM, Skoops wrote:
At 11/27/25 03:21 PM, Dolorious wrote:
At 11/27/25 04:33 AM, Skoops wrote:I've been asked to implement it in my workflow at my job, I learned as much as I could stomach since I was being paid to do so. Ultimately, as someone that was already competent at my job, it didn't make anything easier or faster, it just crammed a slot machine in the middle of my workflow, a major timesink to get something different than I imagined. It was always the bottleneck, not the "tool" people keep insisting it must be. After my trial period, I told my bosses what I already knew: it's not worth it. Can't make anything that would be made public because everyone would clock it as AI and it would damage the brand, can't use it for internals with clients because it makes promises in preproduction that we can't keep in production. Better to have sketchy and not-so-polished designs that we understand vs something that looks perfectly rendered at a distance but makes no fuckin sense up close.

That must've been a pain in the ass trying to work around that shit. I imagine it's like working with a co-worker that you have to fix all the fuck ups for, but can't exactly teach to not make the same mistakes. Businesses going all in on that is going to backfire. Hell it already has when Valve just recently caught Blackrock's lawyers using AI in a legal battle. The stupid machine hallucinated points and evidence that didn't exist.

It's something like that, yeah. It's like having an employee that you have to train from the ground up every time there's a new task, only he doesn't understand your language.

I used ComfyUI for image generation, which is supposed to be the most fine-tuneable shell for Stable Diffusion, and I guess it is. It requires heaps of plugins and tangled webs of node-based instructions to have any chance of getting what you "want," especially if you need one thing to look a particular way and another thing to look a different particular way in the same image. You can't just bash different generations together because they change to a different random angle and perspective every time unless you're doing a "controlnet" which is a whole can of worms in and of itself. Suffice to say, you'd need to be most of the way done with doing something by hand if you wanted it to "help."

So it takes a few hours to get set up and another 30+ minutes of pulling a slot machine handle before you get something that's close enough. You'd think that once you're all set up with the node layout and everything that you wouldn't have to do it again, but nope, every time you gotta get a new type of image, 20% of that setup needs to be changed. There's never a point where it gets easier. The knowledge ceiling is fairly low, but the fact that it sometimes randomly does what you tell it to do (despite all the reasons it can't do that consistently) makes people think that the knowledge base is deep and you can wrestle its limitations into submission with enough custom nodes and whatever. It's amazing how superstitious and ritualistic the people that actually like this shit are; they've all got their own little magic tricks and folk remedies that they swear by that don't actually change the odds at all.

All that to generate a base for what I'd then have to work on manually.

So, like, at this point I know more about how to use it than anyone at my company despite being the one that hates it on principle more than anyone. My bosses still think it's a magic free labor machine, and I do what I can to dispel that delusion. They only ever see the successes on the internet and assume people are getting those on the first try, so I keep all the failed generations and time logs saved so I can show them how many failures it takes and how much time it wastes trying to get something close enough which, again, does us no favors in actual production. Of course, I can't get through to them all the way. They think whatever hard limitations it has now will just magically go away someday soon, but as someone that's spent enough time with it, I'm pretty sure those things are baked in. It feels a lot less scary when you actually experience just how much you have to fight against it to get anything that isn't broken, let alone what you asked for.

Yeah sounds about right. Heard very identical stories from programmers/coders who were told to use AI to generate code for whatever software from their boss. Having to go through the slot machine process, then spend many hours fixing all the fuck ups the LLM generated. Even coders are annoyed with this shit.
But yeah LLM's are schizo as you can tell with it's randomness and it's concerning that doctors, lawyers and other important professions are starting to rely on it.

Heard a recent story of some guy using AI to try to identify edible mushrooms. Man uses the machine to examine a mushroom. Machine says the mushroom is safe. Man eats the mushroom. Gets fucking poisoned. I doubt we will have to worry about anything like Skynet. The closest we're going to get is the Cyberpunk 2077 AI's. The ones that act like full schizo viruses. Except it would lean more towards envisioned Y2K style of fuckery.


You know as fucked up and scary as a Cyberpunk 2077 esque future with Y2K style fuckery sounds like I would be mildly entertained by such a future in a morbid way.


@Skoops

Reading about your job experience with that technology was very insightful and only confirms what I suspected, thanks for typing it out.


At 11/28/25 12:10 AM, Dolorious wrote:Heard a recent story of some guy using AI to try to identify edible mushrooms. Man uses the machine to examine a mushroom. Machine says the mushroom is safe. Man eats the mushroom. Gets fucking poisoned.

I remember something like this being one of the earliest AI scandals, with an AI mushroom-identifying book available on Amazon that had blatantly wrong information.


BBS Signature

I remember seeing a post on another website the other day, and they made a good point.


The people that had to fight tooth and nail with their portfolios to get the job that they wanted, learning fundamentals and other art studies topics.


Only for some to lose their jobs over AI that is very noticeable, and what's even worse is when they DON'T clean up the mess that it could leave and just post it publicly.


There's a very good reason why people are against it, and I can't say that I don't blame them.


I don't use them in my workflow. Skoops already did a good job explaining it and I personally just find it super ugly. I go by my gut feeling when something is AI or not.


I find it hilarious that the same companies hiring the artists they just fired to clean up the animations. Just goes to show that AI isn't as glamorous as CEOs and other companies make it out to be.


I have seen some instances where programmers use AI to figure out programming troubleshooting or users having to use AI to generate a code for their projects, and I find that rather... unfortunate. Though coding is more brutal.


I have been using ChatGPT to help me with production and scheduling tasks, but nothing on the actual art or creative side.


It's been helpful for someone like me with a history of being super unorganized.


It is genuinely so fucking sad to see so many people sacrificing their own skills and talents to a soulless machine. If you can't make it, then learn how to like everyone else has!


BBS Signature

I'm predominantly a game maker now and I've seen a lot of ppl using it for RPG npc characters for radiant scripts and whatnot. Im not making that type of game but ppl who do tween animation or bone rig 2d (or 3d) have been using 'AI' for years. Sure they're just smart algorithmis but if you ever closed a shape with point nodes in blender or adobe you're using 'ai' - same with auto color fills etc.I wasn't talking about full - let me generate some video and post it as a game or animation. I meant actual AI tool that are being added to adobe, paint tools, blender etc. Not all of these are pure generative.For example color matching, motion tools etc are AI algorithms.


At 11/28/25 09:35 PM, Kel-chan wrote:I'm predominantly a game maker now and I've seen a lot of ppl using it for RPG npc characters for radiant scripts and whatnot. Im not making that type of game but ppl who do tween animation or bone rig 2d (or 3d) have been using 'AI' for years. Sure they're just smart algorithmis but if you ever closed a shape with point nodes in blender or adobe you're using 'ai' - same with auto color fills etc.I wasn't talking about full - let me generate some video and post it as a game or animation. I meant actual AI tool that are being added to adobe, paint tools, blender etc. Not all of these are pure generative.For example color matching, motion tools etc are AI algorithms.


I wouldn't even consider tweening or interpolation the same as LLM's being used. The LLM's being fed constant data from various sources, that you just type in whatever prompt and it does absolutely everything for you as if you commissioned somebody to make the whole ass project. Meanwhile basic interpolation you still have to manually go move, deform and adjust the speed for whatever 3d object you're animating. Same type of deal with 2d drawing utilizing puppet rigging when you're rotating a limb or something. At the end of the day you still have basically have full control where you're manually drawing or moving objects. Nobody called that shit AI for many years. The same way nobody called any digital tool with full or still near full control AI until recently.


Now I'm not sure about auto color fills as I don't know much about it. Unless you're talking about paint bucket fills.

Anyways when AI is talked about. Nobody is talking about basic tools that's been used by animators and artist for nearly decades. They're talking about those LLM's that generate deepfakes, entire paintings and full length animations. The things where the only human input is some simple text prompt for it to interpret to generate slop.

Hell these LLM's aren't even true AI. But they're being marketed as that as companies are going all in on these.

Shoving what is basically RNG into everything. That's a Y2K type of event waiting to happen. Machine roulette on whether it goes full schizo and or bricks itself and everything it was implemented in when performing tasks. Assuming they get that far with the scam.



Heard it has been used for job applicants or telemarketing machines. Work that is to be emotionally distant. Get results quickly and the methods are irrelevant. Work that look for any exploi... I mean key word finding and are now being program to blackmail. Would you accept criminal simulation?


Something that came up as I was browsing social media lately:

iu_1504750_3993712.webp


Upon firther digging there are several articles but here is one for further reading.


I love AI! I'm trying to learn to make AI, and AI is generally a cool thing to experience. It's a great tool when used correctly.

"when it's used correctly."

People use it for AI slop, brainrot, and other things they are too lazy to make themselves. It's not creative, it's trash, and when people lean on it, it disgusts me. It is not okay to try to make entire stories, animations, games, anything effortlessly and pretend you actually did something.

AI is a tool. Use it, don't depend on it. If it's taken away from you, you shouldn't be helpless.

And, yeah, that's my opinion on AI. I use it sometimes to fix bugs in my code. That's basically it.


I love smiley faces :D


I love coming up with cool ideas but I'm terrible at completely fleshing them out... most of the time I just get other people and friends to help me with basic ideas and lore.... but... sometimes... for the big ideas... I go to chat gpt and uhhhhh put it in and... uhhhhhhhh yeah, it does help... a lot... because I can come up with the main cool idea and then I just ask "hey filthy clanker... give me 5 more versions of this and keep fleshing it out"


Honestly, I do that too sometimes. This is a perfect example for using AI as a TOOL, but not leaning on it and using it to make entire franchises.

It's a bit lazy, but still, AI is there to help. Just don't abuse it.


I love smiley faces :D


in my humble opinion, I love the talent of newer animators but I miss the good all days of flash

this is why am I opinion we need to. . .


MAKE FLASH POSTING GREAT AGAIN!!!


(Yes, I know it’s AI and I know this is Trump like him or hate him, but I just found this on some random website and I thought it was really hilarious)

iu_1505890_19559243.jpg