Making Images in the Cloud vs Local

Having been a Professional Designer [Architect] for 50 years I will never adopt AI as a 'tool' for graphic art, not when I was doing life drawing 54 years ago.

Architecture itself has been rendered mechanical and lifeless by CAD already, even before the 1-click wonder of synthetic craft.

Having Used Both Local And Site/Web Based AI Text To Image Generations, I Think Its Here To Stay, It Will Only Get Bigger And bigger As Times Move on. I Understand Those Like Jafo Who Say Its Not For Them As They May Have Spent A Lifetime, Learning, Training How To Do Professional Drawings From Scratch. 

In Time I Think AI Is Going To Become "A New Normal Way Of Doing Things" And This Should Annoy And Anger Those Who Have Spent A Lifetime As Pros Doing Things In The Way They Were Educated To. Some Professions Already Heavily Rely on AI, More And More Will Follow Suit As Time Passes. I Think It Will Come Down to "What Jobs Can Or Will AI Take Over", So This Will Always Be A Thorn In Some Sides As It Gets Better And better At Doing Things That Once Took Great Learning Curves And Years Of Training To Do In The Past.

I Only Adopted AI As I Was Not Happy With Lots Of The Walls And backgrounds I Was Trying To Match The Windowblinds And Themes I Was Using To Customize My Set Up, It Still Drives Me On, But Now Its Having Medical Advantages In Slowing My Vascular Dementia, So At Least In My Case, Its Now "A Must Do Thing".

I Had No Idea Just How Popular AI Text To Image Would Become With Other Users Who Like Similar Things To Me, It Took Another User To Tell Me How Many Views A Topic I Started "AI Wallpaper/ Art" Last Year Would Become, With Over three Quarters Of A Million Views, So Much So, I Was Asked To Create A Monthly Topic For The Pieces I Create As The Main Topic Had Become Too Large, With Too Many Pages To Navigate Easily By All The Users Viewing It.

Good, Bad Or Indifferent, AI Is Going To Be Incorporated Into So Many Future Things We All Do, Is That Good Or Bad, Depends On The User, For Me, Its Now A Must And I Enjoy The Process Of Creating Walls That I am Happy To Match With The Theme Or Windowblinds I am Using On Any Day Or Week.

I change my desktop background almost every day. I like the variety of images available on WinCustomize, most of which are AI generated. The only criticism I have is that AI seems to have a very hard time with human hands, building roofs and windows, and miscellaneous other items. While I really appreciate all the excellent work by most "artists" on WinCustomize, I wish more time was spent on post generation work. Thanks to everyone creating all the awesome images.  

Yes, that was me who prompted naroon1 to post his monthly AI graphics thread.

It may not be something I will ever use, or want to use, but I know it's the future and is and/or will be extremely popular.

I only know that because of my history of graphics without its use.

Only answering partially, here.

My problems with AI are that it's too easy and too polluting. Too easy because formulating prompts isn't all that hard, one can even use AI to help one do it. I spent a good deal of time trying and (only slightly) succeeding learning a bunch of programs and techniques, etc. and a lot of it feels wasted.

I can tolerate AI's hand/perspective problems because correcting them isn't inordinately difficult.

What I find frightening isn't people/machines creating, it's the cost...bot monetary and ecological. Leaving the money aside, it's been estimated that to catch up with AL'S energy demands, 50 gigawatt range atomic reactors are needed. That's just for now. That will only increase.

Frankly, I'm rather frightened by that, and I freely admit it dwarfs my petty concerns about the future of art which, I suspect, will survive.

Quoting pelaird, reply 3

I change my desktop background almost every day. I like the variety of images available on WinCustomize, most of which are AI generated. The only criticism I have is that AI seems to have a very hard time with human hands, building roofs and windows, and miscellaneous other items. While I really appreciate all the excellent work by most "artists" on WinCustomize, I wish more time was spent on post generation work. Thanks to everyone creating all the awesome images.  

I Understand In The Past I have Posted Creations with Finger Issues And Possibly Many Other Imperfections. These Issues have been dealt with By Introducing New Models And Renderers, You May Not Have Seen These Improvements In The Bodies Of Work I Have Created As Due To My Dementia, I Tend To Stick To What I Like and Have Used In The Past. Even "Dementia Man" Can Evolve, So I Too Have Started To Immerse Myself In New Models And Renderers.

Sites That Offer "Text To Image" Creations have always been aware That Not All Models Have The Same Strengths, Some Try To Be "Entry Level", Giving Generally Good Results Overall, But With Some Issues. Some Models Have Specific Strengths Like, Good With Faces, Or Good With Written Text Etc. In The Couple Of Years I Have Been Creating Pieces, The Site I Have Used, Have Introduced Dozens Of New Models, Some Giving An Overall Adherence To Your Text Input , And Many More With Specific Uses For Creations. The System Works On A Credit System, You Can Either Buy Or Win Credits, I Have Done Both. Newer Models Tend To Cost More To Use And To Apply After Effects (As They Should, Afterall , Business Is Business). I Suppose Its Down To The Person Which Models They Choose, Usually Based On How Many Credits They Have.

Example Z Image Turbo Costs Around 1 to 3 credits to create your image, Then More Credits To Use Creative Upscale(12) But Can Give Imperfections As Its Cheap To Use.

GPT Image High, Costs 22 Credits Per Image And Then More Again To Use Upscale Features, But Gives More Accurate Results Including Hands And Fingers And Accurate Buildings Etc..

So Getting The Closest Results To What You Type Costs A Lot More. Its Never Going To Be 100% Perfect But A User Can Get Much Better Results If They Pay A Lot More. 

 I Predict A Future Were Text To Image Models Become So Good, Most People Could Never Tell The What Is A Real Image And What Is An AI Creation, And I Think That Future Is Not Too Far Off.

Thanks for that explanation naroon1. I have noticed your images are much cleaner lately. I will  usually clean up the worst abnormalities in Photoshop of images that I want to use on my desktop. I have a 40” 4K monitor, so any imperfections really stand out.  

I very much enjoy local AI both for art and for text generation. I've been wanting to get away from the cloud as much as possible because it keeps data as our own and it can handle doing a lot of things at a much cheaper price.

This is a great area for Stardock in my opinion. I think as cloud (tokens) and api's continue to get more and more expensive the demand for local AI will increase and I would love to see more people get into it. If for nothing else to increase the demand for local gpus so big companies and data centers stop hopping them all 😅

I only have a 3060 gpu currently BTW and it can handle more local AI then many would expect. Although, I'd prefer to upgrade to a 5070ti at some point

I Use a 26" 2K Monitor At My Desk But Have A Hdmi Line Going From My Graphics Card To My Second Monitor (Well Actually Its A 4K TV, Also a 50" Like Your Own).

Many Times I've Made A Wall At My Desktop That May Have Used Creative Upscale, Then Later When I Start To Watch Something Back From My PC To The Big Telly, I See Many Strange Things In Such Clarity. As I Have Previously Explained, "Creative Upscale" Can Hallucinate Image And Background Details Based On Yor Prompt, I Did Not Notice As Much On The Smaller Screen, Some Creative Upscales Can Be Fantastic, Adding To A Feel Of A Creation, Some Not So Much.

At The End Of The Day, Like A Lot Of Things, It Comes Down To Money And The Cost Of Credits. I Can Spend 50 Credits For 1 Image Using The Best Models Which 99 Times Out Of 100 Will Give Perfect Results, But Can Also Create 5 or 10 Very Good Creations For The Same Cost. If A User Stuck To Only Using The Best Models It Can Burn Through Hundreds If Not Thousands Of Credits Very Quickly Indeed.

Seeing As 99% Of The Time, I Create AI Art For My Own Use To Match Whatever Theme I Am Using, I am Just Happy To Keep My Mind Busy Using Lower Spec Models As I am Still Learning AI Text To Image, Even After A Few Years.

I am curious what Stardock is doing to make image generation easier on a local setup?