Hacker News Reader

Comment by safog | Hacker News Reader

FAO: People who're claiming GenAi will replace artist jobs.

It might happen but it's extremely difficult even using state of the art models like stable diffusion v6 to get consistent results. There's usually some part or the other of the picture that's broken and it takes a lot of work with prompting, blending, varying to get it to work.

Buy the subscription for 30$ or whatever and give it a shot. First you'll be amazed but then you start noticing the minor flaws and how much effort it takes to make it good enough.

It's still possible that some jobs will get replaced, not every NPC detail needs to be hand drawn perfectly, but let's see.

View on HN

Mini-mills.

AI isn't going to jump in there and do you another Sistine Chapel or Godfather.

You have a restaurant, it needs some random Chinese/Italian/Mexican artwork. Are you going to buy a poster? No way, just generate it and hang it up.

You have a kid who needs to be entertained. Do I get a team of 200 to write me Frozen, world class singers to sing the songs, and animators? No. We'll start with our own 5-minute Peppa Pig style cartoon.

Your widget shop needs a website. Do you go to Getty and find a shot of happy couples or landscapes taken by a professional photographer? No, nobody cares if your happy family has 7 fingers on each hand. Generate the photo, plonk it on your website.

As AI swallows the cheap, low-risk stuff, it also gains scale to do higher forms of art. Someday your detective story with weird sweaters and cops with a fcked up home life will also become AI generated.

Eventually it will climb mount impossible.

Not unless they stop using copyrighted material it won't.

Remember, the legal cases are still ongoing.

I think that restriction would cut the dataset by 90%.

Didn't Adobe Stock essentially opt everyone in the library in before springing it on them?

Maybe it will, maybe it won't.

It took Amazon a decade to have a serious impact on the economy, even though it was working from day one. Now we have AI technology that's only half working. So what the impact will be and when is anyone's guess.

An artist (illustrator) friend sent me this link (NSFW) of another artist detailing his process of using img2img to stylize/enhance his illustrations: https://www.artstation.com/artwork/yDb1ax

Cheesecake content aside, I love seeing AI image generation used as an aid to an artist rather than replacing the artist.

That lower left arm is much too long in the AI generated rendering. Granted, it’s also a bit too long in the drawing, but the rendering makes it even longer.

My sister owns an etsy shop that sells customized items.

They have a graphic designer who's in photoshop all day doing the same repetitive task for each item that's ordered.

This is the first category of jobs to be killed off. It can't already be killed off because it requires some human judgement (at least right now) to make sure photographs on a custom ornament are centered and cropped in a visually appealing way.

I can see AI killing off that kind of low skill graphic design with very limited judgement calls (or creativity) needed.

Couldn't conventional software have done this? And it didn't. So why would AI do it now?

They're not big enough to pay a developer to write and maintaining a software system.

From what I've heard, they were able to use ChatGPT to write applescript shortcuts to automate a lot of the tedious repetitive tasks, without knowing how to code.

Basic things like "Open file in photoshop" then "mark file as in-progress in google sheet", etc...

So basically the value add was being able to write applescript shortcuts without knowing how to code just by going back and forth with ChatGPT.

(This is a boring example but something us developers might take for granted re: use cases for non-developers)

True, it's good as a one shot generator to then get a bunch of results and decide which to keep. Once you have to modify the generated result you usually also don't have layers or context to easily modify it anymore, so you end up wasting if not more time than if you would've done it yourself. This doesn't apply to everything though and it's good at prototyping and getting a general idea or feel for the thing that you're trying to make.

FAO?

for attention of

https://fluentslang.com/fao-meaning/

For the Attention Of

"For Attention Of" https://dictionary.cambridge.org/us/dictionary/english/fao#

Apparently, something that can be replaced with the word "For". It's odd, because FAO takes a LOT more text to get it's point across compared to "For" even though technically they are both the same number of letters.

Consider also that "For" sounds natural, but in this case, FAO does not.

I was about to say - https://www.google.com/search?q=FAO

> Food and Agriculture Organization of the United Nations: Home

(Obviously NOT what the original commenter meant, but no context!)

there’s a lot more to game development - environment textures, normal maps, foliage, gui textures, background sounds, concept art, etc and AI has been making a lot inroads to these areas

People over index on the hard cases but those are not the majority of work.

Curious if there are models trained to output 3d stl (or similar) files instead of 2d png

Yes, including an open source model from openAI (sharp-e). It’s not very good but it does do it

I know it won't replace human artists right off the bat but it will help them accelerate their development. Just like coding assistant, artists can use this to visualize their ideas/drafts and then build upon to tune the work.

Tencent fired about 75% of their 3d bone animators last year. All because you could feed the bones to AI and type "it jumps", "it runs", "it crouches" etc.

WTF is Stable Diffusion v6? SDXL is latest and before that 2.1 lol

Probably reffering to midjourney version 6

Reply on news.ycombinator.com