-
I want to thank all the members that have upgraded your accounts. I truly appreciate your support of the site monetarily. Supporting the site keeps this site up and running as a lot of work daily goes on behind the scenes. Click to Support Signs101 ...
the ai vector issue
- Thread starter Geneva Olson
- Start date
Boudica
I'm here for Educational Purposes
Adobe can create text to vector files... I don't recall which conversation your thinking of. Lemme try a search real quick
Was it this thread?
be prepared for Ai generated artwork
So Illustrator has updated their software today and there is an emphasis on ai generated artwork. You can draw a box on the artboard and type in pretty much anything and it will generate vector art for you. I can just picture the oddball things you may receive I typed in frog in a pickup...
www.signs101.com
kcollinsdesign
Old member
AI can create vector images. The question is, "Is it good enough?"
It's like drawing with a pencil or drawing with a Rapid-o-graph. Pencils give you an indication of the line buried in the deposits of graphite. The brain picks up where the accuracy of the media fails. For many this frees up the drawing process and can lead to more expressive drawing. A Rapid-o-graph produces a more defined line, and can lead to tighter drawing and more restricted expression.
Creating art using vectors is most successful when the drawings are geometric. AI can do this all day. But when the drawing is more expressive and less defined (like pencil drawings), AI struggles to find the intent and uses ones and zeroes to come up with a translation rather than human judgement. A good human vectorizer will understand the artists intent and using their advanced skills create a translation of the artist's intent. It may not be an exact copy, but it will be a more sophisticated form that will likely communicate better with the intended audience.
An interesting parallel is traditional comic book creation. From initial storyboards come "pencils", which then goes to inkers for the final camera-ready art. Each stage involves translation and judgement. Programs like "copainter" have a long way to go before they catch up to someone like Joe Sinnott!
It's like drawing with a pencil or drawing with a Rapid-o-graph. Pencils give you an indication of the line buried in the deposits of graphite. The brain picks up where the accuracy of the media fails. For many this frees up the drawing process and can lead to more expressive drawing. A Rapid-o-graph produces a more defined line, and can lead to tighter drawing and more restricted expression.
Creating art using vectors is most successful when the drawings are geometric. AI can do this all day. But when the drawing is more expressive and less defined (like pencil drawings), AI struggles to find the intent and uses ones and zeroes to come up with a translation rather than human judgement. A good human vectorizer will understand the artists intent and using their advanced skills create a translation of the artist's intent. It may not be an exact copy, but it will be a more sophisticated form that will likely communicate better with the intended audience.
An interesting parallel is traditional comic book creation. From initial storyboards come "pencils", which then goes to inkers for the final camera-ready art. Each stage involves translation and judgement. Programs like "copainter" have a long way to go before they catch up to someone like Joe Sinnott!
WildWestDesigns
Active Member
What no love for the colorist and the letterer? Time to put up my Dr. Ph Martins and Ames guide. Now for cartoon strips, it could all be the same person though.
In the case of the Text to Vector feature in Adobe Illustrator, the function actually works from raster/pixel-based imagery and then automatically auto-traces the result. You get results that aren't very clean. The stuff isn't production ready. And that's just judging the line work.
Generally speaking, so-called AI isn't very good at generating abstract graphics. The various tools do not deliver clean, production-ready elements. At best, the tools are only good for brain-storming/ideation. To me, the AI generation tools feel like using a slot machine. You click a button and don't know what you're going to get -even if you're re-using the same text prompt. It's not predictable. It's not consistent. The results are often riddled with strange errors.
Over the past year I've seen a fairly big increase in the number of customer provided "logos" that were generated by some kind of AI tool. Every one of them has had to be re-drawn in vector form. But not only that, the vectorizing process has been complicated by the visual errors that had to be corrected. On top of that, any lettering generated is bound to be some mystery meat kind of thing that doesn't exactly match existing fonts. Sometimes I've been able to substitute the AI lettering with something close enough. In a couple other cases I've had to vectorize the letters.
Generally speaking, so-called AI isn't very good at generating abstract graphics. The various tools do not deliver clean, production-ready elements. At best, the tools are only good for brain-storming/ideation. To me, the AI generation tools feel like using a slot machine. You click a button and don't know what you're going to get -even if you're re-using the same text prompt. It's not predictable. It's not consistent. The results are often riddled with strange errors.
Over the past year I've seen a fairly big increase in the number of customer provided "logos" that were generated by some kind of AI tool. Every one of them has had to be re-drawn in vector form. But not only that, the vectorizing process has been complicated by the visual errors that had to be corrected. On top of that, any lettering generated is bound to be some mystery meat kind of thing that doesn't exactly match existing fonts. Sometimes I've been able to substitute the AI lettering with something close enough. In a couple other cases I've had to vectorize the letters.
DL Signs
Never go against the family
AI was trained on raster images, not vector, and I doubt it ever will be. Imagine the resources it would take, programming the training algorithms to create something using an unlimited number and style of lines, nodes, colors, blending, effects, layers, masks, and more, and more, and more. Very few people can wrap their heads around vector, much less a bunch of programmers who aren't artists. Since most of what AI generates for users is social media memes, not likely vector is high on their radar anyway.
AI created vectors would be better for our industry, but even large scale raster is a stretch for the so called AI. All it can do is limited by the size/ resolution of everything it was fed to be trained on, which is why it struggles to create normal size images without those "up-scaled" anomalies and appearance flaws. The vector world is where artists, designers, and illustrators can do things it can't. Much like how you don't see any advancements with AI into the engineering world, like vector there are too many variables and rules that programmers cant account for, and lower number of users to make it profitable when they drop the hammer and start trying to recoup the cost. We can't forget that pesky truth that AI can't actually think, or create anything new. If getting a vector created in AI requires the author to do it, there has to be a reason. Might be insightful to find out why.
There's not an AI or trace program that can come close to doing the kind of stuff we can do in all vector. Ain't happening in our lifetime either. In fact, AI technology as a whole has pretty much maxed out, and hit a brick wall. Until we have some pretty serious technological advancements, it's stalled. The hype is still there to keep investors, but we haven't, and we're not going to see much major improvement until then, which is why so many are concerned that it's all going to collapse. So lets all keep doing our vectors and thumbing our noses at all the trillions of dollars people have invested to not even come close to doing what even an old reprobate like me is capable of, and AI isn't.
Hell, I do these for fun... Go ahead, pick your favorite AI and ask it for an editable full vector like this, and see what you get.
Cheers, and Happy New Year!
AI created vectors would be better for our industry, but even large scale raster is a stretch for the so called AI. All it can do is limited by the size/ resolution of everything it was fed to be trained on, which is why it struggles to create normal size images without those "up-scaled" anomalies and appearance flaws. The vector world is where artists, designers, and illustrators can do things it can't. Much like how you don't see any advancements with AI into the engineering world, like vector there are too many variables and rules that programmers cant account for, and lower number of users to make it profitable when they drop the hammer and start trying to recoup the cost. We can't forget that pesky truth that AI can't actually think, or create anything new. If getting a vector created in AI requires the author to do it, there has to be a reason. Might be insightful to find out why.
There's not an AI or trace program that can come close to doing the kind of stuff we can do in all vector. Ain't happening in our lifetime either. In fact, AI technology as a whole has pretty much maxed out, and hit a brick wall. Until we have some pretty serious technological advancements, it's stalled. The hype is still there to keep investors, but we haven't, and we're not going to see much major improvement until then, which is why so many are concerned that it's all going to collapse. So lets all keep doing our vectors and thumbing our noses at all the trillions of dollars people have invested to not even come close to doing what even an old reprobate like me is capable of, and AI isn't.
Hell, I do these for fun... Go ahead, pick your favorite AI and ask it for an editable full vector like this, and see what you get.
Cheers, and Happy New Year!
Attachments
WildWestDesigns
Active Member
I would argue that the hype is the only thing keeping it from collapsing. It is otherwise at that point. It's already had just about everything (if not everything, even if it was in an unofficial capacity) as it is and I would imagine that it's already training on itself (the collapse that you referencing).
Really the only thing that is keeping it going is investors (hoping that they can keep on investing out of this situation before they are unable to stall) and those that are not capable of doing anything without this type of "help". While there are some people using it as a "tool" at what level and is it at a level that couldn't be done with just photobashing using one's favorite search engine or if they are a developer, are they using it as a fancy LSP.
, just at a boring capacity, but as how "AI" is done now, I would argue that it's lackluster.
I have no doubt that it will still be here as too many companies have too much invested into it (why people are getting "AI" Windoze), so to keep from having too much egg on their faces, have to still utilize it. But in the end, it will just be a handful of companies with, maybe even slightly less unique models, which in of itself will make things bland, boring and safe (which companies love). In the end, it'll probably just be auto conversion 2.0 as far as artists are concerned and LSP 2.0 for devs. Brave new world out there.
pro-UP
New Member
It's interesting to use for brainstorming and text ideas, but the control necessary does not exist. It's exceptionally hard to reproduce repeatable graphics. It brings to mind what I learned in statistics with the 50-50 90 rule. With the flip of a coin you have a 50-50 chance with a 90% likelihood of being wrong. Each flip of the coin resets the odds. Each graphic and artistic request is a new coin flip and you never quite know what you will get. It's an aggregator. The clamor for the ever elusive AGI is something that doesn't seem possible and will not exist with the current technology. They have scoured the internet, reviewed and been fed countless works with no consideration for copyright issues, and will eventually hit a wall. The idea behind agi is that it will eventually replace human intelligence and be self-operating and thinking, capable of decision making independent from any programming. The amount of data necessary to come even close to that, considering what has already been gobbled up, is just not possible and cannot exist.
There is no question it's an interesting technology and kinda fun to play around with. We are working on integrating some ai features into our workflow and custom software we are developing (in a highly contained manor). I see value in various applications, but I do not see it replacing the specialty necessary to produce layouts that can actually be fabricated. It will require extensive training. Who will do that? Sign companies? Sign designers? Signage companies highly value their art and are protective of their individual processes and unique way of doing things. There is also the security issues intrinsic to ai. I saw an article recently on how a security specialist tricked the internal Microsoft chat system to give him unfettered access. I think this was the article from Tech Republic.
DL Signs
Never go against the family
I agree, it'll stick around, but as what? Will it stay the course?
With investments at risk as advancements slow to a crawl, Open AI has already signed deals with Walmart to do in app purchases. Watching ads for CoPilot laptops plug how it can identify products in images and videos is just a little too convenient, Meta being nothing more than a massive AdBot, browsers going deeper down the ad-hole every day, all tied to, and under the guise of AI. This seems to be what happens to everything today, if it can host an ad... It will.
So I have to ask if this will be any different. Will AI technology really advance to be a useful tool some day, or just end up a massive marketing/ shopping machine that chats with you and does poor artwork on the side? Who knows...
hybriddesign
owner Hybrid Design
for simpler logos vectorizer.ai works well. It’s $10 or something per month and we use it daily. I usually try illustrator’s auto trace first since I’m likely already working in that program but vectorizer is a lot better at it than adobe. Our designer says it’s the best $10 a month we’ve ever spent
somcalmetim
New Member
Yep Vectorizer.ai is very good...has good color grouping options you can play with and it makes better vectors than any "autotrace" ive ever seen before...not just a bunch of nodes and 3000 colors, it gives fairly nice well formed vectors if you give it half decent artwork...
Also super surprised by Topaz Gigapixel AI for making high res photos out of crap...it doesnt just add resolution it actually adds detail back in...if you let it get too creative it will even add shit that was never there...
Also super surprised by Topaz Gigapixel AI for making high res photos out of crap...it doesnt just add resolution it actually adds detail back in...if you let it get too creative it will even add shit that was never there...
pro-UP
New Member
There was an article I read that said they are going to integrate ads into ai chat responses (although this was in 2024, so I guess they are still working on it, lol).
WildWestDesigns
Active Member
That's the big question. There is all this hype, some people are impressed (and for most normies, I can understand it). Now, it does fine when everything works within certain parameters, just like auto conversion of old does. The simpler the original design, the better of a chance that it has of correctly churning out something that could work. Where it doesn't work within those predefined situations is where you get the lackluster results and those lackluster results are still the majority of the case and I don't know if how it's implemented now, that they can be corrected.
While serving ads is what we are easily seeing on our end, most of this is about data collection. That is were the real money is. It's being sold to us as some type of convenience, efficiency feature, but it's not really for us. We are not their customers anymore. It's the ones that can use that data. Recall, which they still put back in, has no functional reason to exist for the average end user. Now, for those in the corporate world, I can think of quite a few advantages with having something like that.
At the very least, I don't really see this being as exciting after all of the hype truly does die down. It usually is the case when we settle into tech, but I think the hype has been done so much for this, I just see it settle even more so down into ho hum doldrums
MS has already done so. Given how much ads are put into the Windows OS as it is, I cannot imagine a timeline where they hadn't do that.
Article
Blog from Early 2025
DL Signs said:
The main reason people are concerned about the AI bubble popping is the financial math ain't mathing. The dozens of circular deals going on between OpenAI, NVidia, Oracle, Microsoft and others (along with some very screwy accounting) can't hide the fact this stuff isn't generating anywhere near enough revenue to pay for all the data centers and other crap related to AI. OpenAI needs to make roughly $2 trillion within the next 3-5 years to make good on its promises. The company made $60 billion last year.
These companies are burning through hundreds of billions of dollars chasing this AI stuff. I loved one computing industry guy's take on all the impulsive spending: "with this much shit there has to be a pony underneath it."
Last edited:
DL Signs
Never go against the family
That circular financing is often a sign that the end is near. The chasing of their own tail in a last ditch effort to buy time and try to secure more outside investors. I'm sure most of us were around when the .com bubble popped, and those were small numbers by comparison, if this bubble pops, it'll be devastating to the entire economy. I'm sure the so called AI will still exist when the dust settles, but will most likely be a more realistic version than the current hype is making it out to be, I'm thinking more like the description below that WildWestDesign envisions. And it still won't do vector art...
These companies are passing around the same pile of money from one company to the next, back and forth, yet counting it on the books as actual revenue. The only bright side to this situation is, unlike so many companies in the dot-com bubble 25 or so years ago, they're not doing it with a lot of borrowed money. At least some of the companies involved can provide a financial back-stop. It's not like some start-up burning through cash provided by lots of mom and pop investors.
Still, the Dow Jones Industrial Average, S&P 500, NASDAQ, etc. have been propped up a great deal by those "magnificent ten" tech companies. A popped AI bubble could take the DJIA down from its current 48,000 level to less than 30,000. There are other factors in our economy that could cause a big sell-off in stocks, such as housing industry price bubble. Housing prices in so many areas of the country have become unmoored from the reality of local income levels. Consumer debt is at all time highs and worsening. The general public is increasingly angry about high prices of basic things like food. There is a growing backlash against data center projects. They're causing electricity bills to rise or even spike in some localized cases. Data centers consume a lot of water. People in my town are pissed off enough as it is about water bill price hikes. They're really not going to like it when a "modest size" 100 megawatt data center gets built nearby and causes yet another price hike. All in all, it's not a good situation.
The risk being posed to the stock market is beginning to create a "too big to fail" situation. Executives at some of these tech companies are now openly talking about "public-private partnership" to help them reach their goals with AI. Translation: when the bubble pops the tech companies will expect a bail-out from Uncle Sam.
Still, the Dow Jones Industrial Average, S&P 500, NASDAQ, etc. have been propped up a great deal by those "magnificent ten" tech companies. A popped AI bubble could take the DJIA down from its current 48,000 level to less than 30,000. There are other factors in our economy that could cause a big sell-off in stocks, such as housing industry price bubble. Housing prices in so many areas of the country have become unmoored from the reality of local income levels. Consumer debt is at all time highs and worsening. The general public is increasingly angry about high prices of basic things like food. There is a growing backlash against data center projects. They're causing electricity bills to rise or even spike in some localized cases. Data centers consume a lot of water. People in my town are pissed off enough as it is about water bill price hikes. They're really not going to like it when a "modest size" 100 megawatt data center gets built nearby and causes yet another price hike. All in all, it's not a good situation.
The risk being posed to the stock market is beginning to create a "too big to fail" situation. Executives at some of these tech companies are now openly talking about "public-private partnership" to help them reach their goals with AI. Translation: when the bubble pops the tech companies will expect a bail-out from Uncle Sam.
Last edited:
WildWestDesigns
Active Member
Not quite there with happened happened in 2008. Closer to 2020 though. Unfortunately, that helping was not a good investment in my mind (they rarely are as the same industries seem to always need bailouts).
Short term investor is really going to be what gets hit. And unfortunately, with how people were taught to handle things, that's no bueno for most people as we have gotten away from doing things and expecting certain things to always be there and always work. Especially when wanting the government to handle the inner workings of it all.
Not surprising. In my lifetime, there have been 8 bailouts provided by the government for businesses/industries that were considered "too big to fail" (just lifetime, not all are within my memory of having to have dealt with). A lot of crossover in quite a few of those bailouts as to who gets them. I am not a bailout fan of corporations. I don't like the government involved in that type of thing as that usually rife with fraud and bureaucracy (usually they go hand in hand). There is a lot of things that I'm not liking government involved in. In the long run, it always screws the plebs (ironically even the ones that want the bailout, but I digress).