Author Topic: Adobe Stock livestream: generative AI stock 101 + live Q&A Wed, June 28 @12pm PT (Read 7559 times)

Jo Ann Snover · « **Reply #1 on:** June 27, 2023, 13:23 »

Perhaps June 28th?

cobalt · « **Reply #3 on:** June 27, 2023, 14:50 »

Will it be recorded for thos of us who want to watch later?

cobalt · « **Reply #6 on:** June 29, 2023, 16:37 »

That was great, thank you very much!!!

ETA

I do have one question - when you do these photorealistic macro insect shots, do you compare them to real photos to make sure it looks authentic?

Because ai can be very creative in inventing new species.

Thank you for the great prompt ideas to experiment with

MatHayward · « **Reply #7 on:** June 30, 2023, 10:11 »

This is an excellent question. It's important you check the AI box before submitting to clearly identify the content as generative AI for customers.

I don't compare with actual images, but I do watch out for obvious issues. For example, I've generated macro images of bees that look fantastic, but upon closer inspection I see they only have one or possibly three antenna instead of two. Those go to the trash bin and I keep trying.

-Mat Hayward

cobalt · « **Reply #8 on:** June 30, 2023, 14:02 »

Currently trying my luck at underwater images. But I am no marine life expert and I am suspecting that the ai is being very creative and inventing new species...so I am going more towards watercolor or designs that are clearly art.

I think this can also be a problem if you are trying to create landscapes with a localized style. That cactus might not actually grow in Mexico...

The customers will learn to pay attention to this, but it is important to consider when creating content.

gnirtS · « **Reply #9 on:** June 30, 2023, 23:36 »

Two of us here are underwater photographers and we spent a hilarious 5 hours laughing at the horrific species, previously unknown to science it invented when prompted for common creatures.

Bottle-nose Orca with penguin feet when prompted for "common dolphin" was one. Literally an entire evenings entertainment as it failed to produce a single, remotely usable item.

cobalt · « **Reply #10 on:** July 01, 2023, 02:04 »

The results were so weird, that even I could see it. Crabs with far too many legs, weird fish, Octopus with 20 arms.

If you choose an „artsy“ design then that is ok, it can even enhance the concept, but it cannot be used for photorealism.

With people or cats or dogs we can immediately understand what is wrong. But other animals or plants?

I am sure also the corals or plants underwater were wrong or mixed together from different parts of the ocean.

I now have a few concept images with dolphins, seals, whales, but even these animals look a little strange. The size of the eyes, shape of fins.

But I am now using mostly watercolor, painting or design styles, so it is very obvious this is not a real photo, or an attempt at being a real photo.

Also a clear color scheme for the series that does not exist in nature.

It is the same problem people encounter with chat gpt. The ai can write great text, but nobody knows if the actual facts are real.

Biggest danger is that people ask it medical questions and cannot see that the answer is completely wrong.

On the other hand…this keeps us fed. As long as experts are still needed to work with ai, it stays a tool that enhances professional use.

It certainly saves a lot of time, I have cycled through hundreds of concepts in 3 days and gotten a lot of interesting results to explore further.

eta

Also got a lot of ideas for photos or videos.So next time at the beach I can look at my ai concepts gone wrong for inspiration.

Jo Ann Snover · « **Reply #11 on:** July 01, 2023, 11:08 »

Mat was very clear in the livestream that genAI images that were wrong in any way, or which included logos or other IP, should not be submitted. That's exactly what should happen, but it is not what is happening.

The inspectors are letting in thousands of nearly-real-but-hopelessly-wrong images every day (I've been monitoring in the past week). So contributors are submitting them and inspectors are approving them.

This has been removed

And this is just a very, very small sample...

PhotoBomb · « **Reply #12 on:** July 01, 2023, 15:24 »

Quote from: cobalt on June 30, 2023, 14:02

Currently trying my luck at underwater images. But I am no marine life expert and I am suspecting that the ai is being very creative and inventing new species...so I am going more towards watercolor or designs that are clearly art.

I think this can also be a problem if you are trying to create landscapes with a localized style. That cactus might not actually grow in Mexico...

Thecustomers will learn to pay attention to this, but it is important to consider when creating content.

I think you give them way too much credit

gnirtS · « **Reply #13 on:** July 01, 2023, 22:52 »

Quote from: Jo Ann Snover on July 01, 2023, 11:08

Mat was very clear in the livestream that genAI images that were wrong in any way, or which included logos or other IP, should not be submitted. That's exactly what should happen, but it is not what is happening.

The inspectors are letting in thousands of nearly-real-but-hopelessly-wrong images every day (I've been monitoring in the past week). So contributors are submitting them and inspectors are approving them.

And this is just a very, very small sample...

Firstly thanks for highlighting a growing problem. But secondly, thanks for providing me with a laugh first thing this morning with those images!

Her Ugliness · « **Reply #14 on:** July 02, 2023, 00:52 »

Quote from: gnirtS on July 01, 2023, 22:52

Firstly thanks for highlighting a growing problem. But secondly, thanks for providing me with a laugh first thing this morning with those images!
[/quote]

Second! The fruit one is my favorite.

But I don't get how these get approved. I understand that sometimes image slook "right enough" at first glance and you do not notice mistakes immediatelly, like with insects that have to many or too little limbs. But everything is wrong about some of these pictures. How can a reviewer not notice this, even if he or she just looked at the image for a second?

Digital · « **Reply #15 on:** July 02, 2023, 03:50 »

Ahaha, I think the correct term should not be "stock AI images", but "laughingstock AI images".

cobalt · « **Reply #16 on:** July 02, 2023, 03:56 »

Many of these mistakes are great content for social media marketing. Endless entertainment value.

Love the hamburger image.

The IP problem is a bigger issue, like with the apple keyboard.

DiscreetDuck · « **Reply #17 on:** July 02, 2023, 04:37 »

Adobe is making a fool of itself here by allowing contributors to feed the image database with all these laughable, pathetic productions. Customers will soon go buy their images elsewhere, at more serious image providers, that's a shame. They will quickly understand the risk of buying visuals that can make a professional publication ridiculous, especially for inconsistencies that would not be visible at first glance.
But maybe Adobe plans to work more with humorous social networks than with communication agencies, that can be a choice, why not. But I remain surprised.

I think that Adobe would have to benefit from closing the accounts of these "producers" of images who do not even control what they drop en masse in revision.
In some time, cleaning the database will be very difficult with these millions of images which are validated without even having been inspected. Adobe will need a lot of volunteers to do this work. Times have changed.

Lowls · « **Reply #18 on:** July 02, 2023, 04:51 »

Adobe is letting itself down in more than one way ...

I have submitted images of gummy bears and Adobe has rejected each for poor focus and yet shutterstock have since accepted and sold most of them and a couple repeatedly. I also have some beautiful autumnal leaf images which are dissimilar enough from the 3 I have been permitted thatbthey should be allowed. And yet rejected for similars. Which then begs the question how is this shower of s£it possible ...

... not only is the image nonsense ... there is a lot of nonsese permitted ... no rejection for similar Matt???
Not only are they similar they're so similar they look like the same.photo. A joke.

Lowls · « **Reply #19 on:** July 02, 2023, 05:00 »

And this ... one can only imagine what their 'Skull' looks like. Its not funny, it's pathetic. Would you like a large fries with your skull ...

Lowls · « **Reply #20 on:** July 02, 2023, 05:15 »

No doubt Matt will say its just teething problems.

A.I. gingivitis vs actual gingivitis

* almighty. What concept teeth did it borrow that off. I wouod suggest that gingivitis is the least of its worries.

cobalt · « **Reply #21 on:** July 02, 2023, 06:34 »

I fully agree on the similars problem, accepting one weird image is fine, but 10 very similar all weird images is not.

I guess those are "teething" problems of the new medium.

However if Adobe starts to reject more, the forums will be filled up again by people complaining they decline too much...

Jo Ann Snover · « **Reply #22 on:** July 02, 2023, 19:26 »

Quote from: Lowls on July 02, 2023, 04:51

...Which then begs the question how is this shower of s£it possible ...

... not only is the image nonsense ... there is a lot of nonsese permitted ... no rejection for similar Matt???
...

I have a collection of examples of massively repetitive similars from genAI images approved in the last week or two.

It's not just a few times it happens. It's many more than 5 or 6 similars.

Even when the images are not "mistakes", when a photographer gets a rejection for similars when there are only 2 or 3 alike, it seems absolutely clear that there's a completely different rule book for the genAI submissions

Here are just a handful from those I've seen
35 yellow suitcases

79 overhead shots of peaches (there are some pie shots I couldn't exclude because the keywords are spammed)

129 sunset on a beach with palm trees

220+ marble wave abstracts

cobalt · « **Reply #23 on:** July 03, 2023, 03:01 »

Maybe with the overload the reviewers just don`t have the time to pick the best files and decline the rest.

So if the batch is technically fine, it is either accept or reject everything.

I am sure in time Adobe will align the review process but for now it looks either they take too much or reject too much.

If the technical quality has a high enough standard, then personally that wouldn‘t be my problem. It is still better content than the ports doing endless fractals.

It might not be fair to other mediums and those of us who make the effort to select well, but if Adobe states this is a temporary issue, I can live with it.

DiscreetDuck · « **Reply #24 on:** July 03, 2023, 03:30 »

Generalized mediocrity, at all levels...
What is mediocre, false, is favored.

MicrostockGroup Sponsors

Author Topic: Adobe Stock livestream: generative AI stock 101 + live Q&A Wed, June 28 @12pm PT (Read 7559 times)

MatHayward

Jo Ann Snover

MatHayward

cobalt

MatHayward

MatHayward

cobalt

MatHayward

cobalt

gnirtS

cobalt

Jo Ann Snover

PhotoBomb

gnirtS

Her Ugliness

Digital

cobalt

DiscreetDuck

Lowls

Lowls

Lowls

cobalt

Jo Ann Snover

cobalt

DiscreetDuck

Related Topics

Sponsors

Microstock Poll Results

Sponsors