Quote from: cascoly on January 02, 2023, 17:28

Quote from: Uncle Pete on January 02, 2023, 14:55

I'll try to make this easier for you to understand. Everyone doesn't own or drive a car. And they are doing just fine. Some people can't afford a car and they are also doing just fine. Plenty of people, that is. They never transitioned into what replaced the horse and buggy.
.

not that it really matters, but

https://edc.nyc/article/new-yorkers-and-their-cars

San Francisco is very car free also. https://sf.streetsblog.org/2014/08/15/car-free-households-are-booming-in-san-francisco/

How do the majority of people who made income from microstock in 2012 do ten years later?

How do the majority of people who made income from microstock in 2012 do ten years later?

Made me curious. About 55% up rolling 12 month total for November 2012 vs November 2022

How do the majority of people who made income from microstock in 2012 do ten years later?

down about 10% since 2012, but next 5 years increased over 2012 by 50%

Do not forget that the dollar inflates. What was the inflation rate in 10 years? Here, downvote it.

There are calculators all over the interent but I went ahead and did it for you. Something over 20% increase adjusted. Of course the last year has been crazy for inflation compared to previous years.

In 10 years the microstock business as we know it will not exist, 90% of the regular clients of the agencies will stop buying subscriptions and will generate their own images using artificial intelligence programs. What is going to happen in the future is that little by little the sales of microstock will disappear and many people will give up and dedicate themselves to something else. One after another the agencies will disappear or declare bankruptcy.

I think ten years is optimistic

https://www.freepik.com/author/dembaxx here's a portfolio made entirely with AI tools. Far down to the oldest "creations", the images look like crap. But the newer ones are increasingly better. Have a look at the captions and keywords, for a good laugh.

... to keep the point, my last post: Food for thought
https://luisafumi-digitalart.com/blog/2022/12/20/food-for-thought/



... and more to come

intriguing post! a great series of blogs about future of AI at https://www.cold-takes.com/most-important-century/

What is the image size and resolution like on some of these AI generators? I one better than the other when it comes to that? DALLE vs Midjourney vs ??

What is the image size and resolution like on some of these AI generators? I one better than the other when it comes to that? DALLE vs Midjourney vs ??

Hi, you got plenty more AI generators but in my opinion the 3 Big at the moment are Dalle, Midjourney and stable diffusion. Also there's a buzz about Playground Ai but i never tried. I am currently using Midjourney subscription and free SD 2.1. via Github/colab.

In terms of Size/ resolution:

1) Dalle- 1024x1024

2) Midjourney v4 : upscale 1024X1024, beta scale 2048x2048 but you loose detail in 2nd
(you can improve detail directly in Prompt using arguments like ultra detail, super detailed, Photorealistic, etc... and request best quality which costs 2 times more)

3a)"Outdated" Disco diffusion / Stable Diffusion 1.5 : 1668x768 with colab pro+
3b)Stable diffusion 2.1 : 768x 768 or 1280x720 (sometimes errors in 2nd)
For both you can increase detail and sharp changing some parameters (example: 25 steps basic, 350 steps more detail, also you consume more power and time)

Note: After... you can upscale with AI tools but you got better results depending the type of image if it is photo realistic, cartoons/anime, 3d render etc...Sometimes it's better to combine 2/3 AI upscale tools but this depends really on the image itself and of course trial error process.

One AI I had not heard about before, Lexica art, built on stable diffusion and returning high end images, check the pricing though. You can begin to understand the 'game changing' talk about this new technology.

https://lexica.art/

Thanks for sharing that, I had not heard about this one as well. Seeing how fast AIs improve  I have been feeling a bit depressed about my microstock future lately, thinking I'd be jobless sooner than I had thought, but this is actually making me feel a bit more optimistic again that it might still take a while till I will be completely replaceable. As always I have been checking my direct competition and the results you get when searching for example for "French Bulldog" are a huge contrast to the amazing images this site shows you on the first page.
 


Not all of the results are as bizzare, but at least among the photorealistic results, none really looks right.
Eventually AI will get there and I will be out of an income, but with any luck I still have some time.

Tony Northrup says AI will eventually replace 95% of stock photography. 

https://www.youtube.com/watch?v=8VRbt8GufeY...

income from existing portfolios will continue - SS would need almost 4 BILLION AI images to form 95% of their offering (assuming zero new images sent in)

the biggest factor will be how many abandon stock altogether, which should remain small for some time

Finn,

I only use lexica to find themes and some useful guides to my prompts that i like. I think Stable Diffusion can be more powerful with trained models or with image references.

I agree, eventually AI will get there... but it has been quite a speed race. I am using my memory here so it may not be precise but in 2022 we had:

-Disco diffusion v3, v4; Stable Diffusion 1.1,1.2,1.3,1.4,1.5,2 and 2.1;
-Midjourney v2,v3,v4a,v4b

I just tried with a simple text prompt of "french bulldog sit in XVII century Throne" and use some technical arguments without image reference (see images). Probably i could tune a little better but the result is ok for at least showing concept to a client before photo shooting. AI is still struggling to learn the teeth, feet and hands since there's still lack of knowledge of anatomy and right positions of any animal.

To be honest i don't know how much time AI will need but it's getting better everyday since it learns concepts, styles and references 24/7.

Yes, these look great, but they look like paintings to me, not like real photos, which was what I was talking about (sorry if I didn't get that across correctly. It was what I meant by "photorealistic", but maybe I used the wrong word). In regards of paintings AI is - sadly - already kicking ass. I feel so sorry for painting artists, even though I know as a photographer I will be in the same boat soon. I have seen some mind blowing AI generated French Bulldog paintings on Adobe. But when it comes to photos it's still struggling, at least for now.

So who is allowing AI Generated art submissions? I think Adobe Stock and Dreamtime are. Others?