about copyright: a new US policy came out that basically says any works generated by AI are considered public domain (not copyrightable). Because copyright only applies to works created by humans. That says a couple of things: 1. any piece of art created by AI can be used by anyone. So most companies/clients won't want AI art as part of their end product or branding. 2. Derivative works from AI generated works, if different enough from the AI work, can be copyrighted (as usual, to standard copyright law). This means artists will be hired to "paint over" or make tweaks to AI art, and AI art will be used to basically rapid prototype ideas and concepts. Note I'm not a lawyer, these are just my findings, and this is just a quick tl;dr on the complex aspect of copyright.

"AI will let us automate labor so we can spend time creating art and music!" was the pitch for AI advancements. But now we're automating a huge part of the process of making art and music so we can spend more time on... what exactly? Creating more rapidfire "content" for corporations and social media?

Everyday this AI thing make me feel that all my hard work in honing my skills in drawing is completely useless

Watching your vid just wanted to make you aware of something. Adobe is not fully transparent about their dataset aside from using adobe stock. Although technically they have the right to use these images, that is only because the opt in the images by default without any option for uses to opt their images out. So not as ethical as they claim. Also Kris is a the person who has been chasing copyright for ai and exposed as a grifter (many chat leaks showing how she is funded to normalise the tech and convince the public not to worry about datascraping issues). Constantly trying to get the USCO to green copyright for prompt base images despite their clear stance after her first attempt.

OK, but there are Midjourney pictures on Adobe stock. So, how exactly do they plan to solve that copyright nightmare?

I don't even get it how are the users even thinking thay they ll be appreciate d for their work when 99 percent of work is done by ai itself

what do you mean - SHOW A CLIENT? how can you show a client something you didn't create and pass it off as your own?

I dumped Adobe when they went to the subscription model. Was one of the first 100 owners of photoshop, but RIP.

I wouldn't call Firefly ethical in any way. The artists uploaded the images to sell them, under general TOS. Years before AI was a thing. Now Adobe took those stock photos and used them in AI, with no warning or any way to opt out or delete those images beforehand. The artists were not recompensed for this their photos being used in Adobe's *private commercial product* . After the backlash, after everything is already done, now Adobe said they will begin to think about giving some pennies to the artists whose work was taken. No way for the artists to have an opinion or opt out from this. Adobe didn't care about the artist's consent or opinions. The artists were not uploading their works to be scraped or used in AI. This is just another case of a huge company screwing artists for profit. Nothing ethical about it.

I want to generate a picture of Brad generating pictures of dogs wearing sweatshirts…on a sweatshirt…worn by a dog.

I should have been doing sports, not creativity. Machines negate the importance of the effort invested in training of drawing and painting.

Isn't Adobe using images posted to Behance, etc? I know Behance mentioned something about art posted being used for AI.

This is evil. Soulless, evil. We work for decades to reach professional level and now we’ll be obsolete in 6 months.

I'm not convinced T2I by itself can save Adobe. As much as the ethical branding does seem to matter, too many startups in the past have succeeded by doing something borderline-illegal first and then patching it up after they get enough capital. For artists who want to focus on art, I sense the most likely path forward will not be with T2I, but to take a sketch they drew themselves, paint in some flat color(easier than ever as of today since Meta put out "Segment Anything Model" which can recognize shapes pretty darn well, even as line art), and I2I it with a customized rendering model. That preserves many of their biggest skillsets - drawing as the interface, then training up a nice renderer(which they can spend a long time on, using whatever methods and materials they like as sources). Many forms of software are going to converge into streamlined interfaces to various models: the distinguishing factor is just the interface, which might be primarily text, or it could be drawing or various forms of live performance. For branding, you could end up with models and prompts that can generate an entire design system... ...or you could see that market reverse direction from marketers building brands to advertise to consumers, towards individuals and communities commissioning personalized "browsing agents" that find people and organizations that service their needs. If the past is a foreign country, we're getting on the boat right now.

Well that's a fast evolution from unrestrained lawlessness to an approach of ethics. It's going to be very gray, though, as it was determined a wholly AI-generated image could not be copyrighted, how many permutations of unattributable work will create enough layers of obfuscation such that an AI could be trained on those images and be technically free of conflict? Or, as the competition for the AI-generation space becomes more heated and there is more incentive to further-occlude the black box of the models, how will the training behavior of competing programs be made accountable, now that ethics IS a concern and primary objective for projects such as Firefly? If a competitor were to, in bad faith, poison a dataset with deliberately copyrighted (or AI generated) input, can an ethically-minded project successfully purge the influence from a dataset? If it ever happened, would they let customers know? Would they not want to know, because it's not about ethics but more about legal exposure? There's an entire new branch of ethics and philosophy being born parallel to this entire industry

With their current dataset, Adobe will never compete. And with their hideous *software rental* model, more and more creators are leaving Adobe for Affinity and other traditionally-priced, non-subscription software. I started on Mac with Adobe everything, but have happily and successfully moved away from both.

This just fuels my depression even more, what's the point in trying to be better when someone can just generate art in seconds

Nice vid ^^ We are not yet in a hyperrealistic ; photographic ; rendering, it "looks" still too digital. A way to fix it?

The logo stuff is pretty cool as it shows that Adobe is likely training this with stock imagery where logos have been modified so that they are more generic and not actual brands, while Midjourneys set has not plastering Coca Cola logos all over. No doubt Adobe knows it can get into trouble with trademarked logos.

Firefly is a dangerous deal.. (think about it)

Has anybody gotten access to use it? I'm still waiting for it

so now its legal to re-sell, because the other platforms kinda shady with their databases.

Can't iterate on selected output? That's pretty limited.

Also all this ai art bs will bite them in the back. If ai can produce the new era of arts and designs, people wont need to buy expensive computers, tablets or ipads. If anything else people will spend the money on a phone.

well here we are 9 months later and I have access to both these programs. Midjourney blows firefly out of the water when it comes to prompt on prompt generation. Though firefly does have a better UI, it feels like its just another stabledefusion based image generator, this ones just built into photoshop.

use --c and a number between 1-100 for more variations in midjourney I recommend no more than --c 20 max in v5, you can also add --s 0-1000 for more stylistic generation 100 is the default I usually use --s 700 --c 10

This is legal, but not ethical. The artists uploading to Stock had the expectation that they would get paid when someone buys their image, now with this AI, all that business is effectively over. Adobe used their work in a new tech they had no idea their work was ever going to be used for, gives them no choice to opt-out their images, and completely killed off all the business they were making by having their images in Stock in order to pull a greater profit for themselves.

I wonder if those self-imposed rules that Firefly has will always make it inferior quality-wise. Since it doesn't have the breadth of learning that other models do.

is this pay to win design?

hi Brad, great stuff as always. I've seen the capability for inpainting and canvas extension but cant find these features, do you know how to access them? keep up the good work!!

Hate this, hate the coverage and the positive PR spin

As far as the US Copyright office is concerned, AI generated imagery is not eligible for copyright. I suppose you could make an argument that the series of prompts you use might be eligible since prompts are are authored by humans, but I don't think anyone has tried that yet. Larger firms will want something that they can retain copyright over and it doesn't seem like this type of imagery is protected. I think there's a case to be made for individual sales from people who just want something cool looking, but for business I think they're going to run into some problems. If Adobe can somehow figure out how to make their generated images copyrightable, game over. They will have built another standard that we'll be using. I wish I shared your optimism. Adobe's intentions seem noble enough. I think eventually we're all going to have to incorporate some aspects of this "synthography" in our work just to keep up. But it doesn't feel good.

More videos 👏👏👏👏

Adobe Stock is outrageously expensive as a service. Amazing quality, but way too expensive for most people. On the other hand, being a member of the creative cloud, I'd make large use of this.

Its been interesting playing around with the Beta. I like how you can set the camera angle in the software.

I really enjoyed this video. You did a very good job of explaining and proposing ideas for the future.

I'm really tired of companies shoving alternative intelligence (I know what ai means ) in my mouth

There is zero ethical argument for using Ai generated art. Unless the artists from Adobe stock get a commission every single time equal to what they’d get if they made it themselves, this is wrong.

The Adobe fantasy pirates look like they're 2D on slightly misplaced parallax layers. AI is going to change the art scene far more than Napster changed music. Adobe chases AI or they're dead.

now to make a movie you just need a script writer😂 And I very excited about the new car models that will be designed through these AI.

There are companies that have bigger portfolios of images than Adobe so Adobe could easily license that if they need a bigger data set. I can definitly see big companies prefer this approach. However small companies and private users will probably be more attracted to free options like Stable Diffusion that you can run on your own hardware.

Short answer: No. In terms of the quality of AI, Adobe is way behind. Everything Adobe is doing now is branding: our images are "legally safe"! How safe? 1. I don't think the quality of the pictures Adobe is generating is a match to its competitors in the market. 2. I don't think Adobe could collect and handle billions of images on the internet. Adobe is a "small data" company (not a "small" data company), it always has difficulty handling big data. 3. Even if Adobe decided to add some copyrighted images in its training set, when the impression of "legally safe AI" is developed, people won't easily challenge that, at least in the short term.

Good video, thank you. For me, AI is nice tool to help in the moments when I'm tired or lazy. Its generative side is helpful to avoid stops that day. We don't need final product made solo by AI but ideas or at least small push in work. My view for now

For these AIs, the more detailed the prompts, the better.

Is it free to use?

I agree that Adobe needs to push and evolve their engine very quickly to catch up and compete. I do love the UI and the diversity of models they plan on releasing as well as the proposed integration with Photoshop.

As an artist. Automátic dislike

Hello

Dude! Dude…😢 So many of the channels, art channels no less, I’ve liked putting positive spins on AI. Take care!

Ugh, now Adobe is on this band wagon

Sounds like something an A.I would say lol

Question, what's going on with your second channel

I think battling over training images is a lost cause, only way to do any sort of style policing if any is at the prompt level... ie "picture of dog in VanGogh style" where its obv the intent was to use someones style...I assume very quickly we are going to have 2nd and 3rd level models trained on models so removed from the original picture that the intent to copy is only even visible back at the prompt level while the lawsuits about training data rages for years to come....even then its sketchy getting a bill from your AI for the Van Gogh prompt and not for your "picture of dog in impressionist style, bright blues, thick oil painting" prompt? I say train on anything you want Weird Science style and then judge the final product like we do now.

All those AI-generated images, Firefly or MJ, all of them look atrocious! No kidding, they look _awful_, I don't know the appeal people see in these. Yes, it's impressive how they are generated, and yes, they can be used as a inspiration, but holy cow do they look off in tons of parts.

just to let you know, I've unsubbed from you now due to these AI posts, they're very disappointing to see

I do not like the propensity of its rendering!!

I hate all the AI. So ridiculous. Im not a fan of people relying on AI to create things, albiet graphics, words or videos. Learn to create on your own.

AI has potential But the way it is now... Well, let's just say the way it is being used and made isn't the best (( from an ethical standpoint ))

Wooow this is amazing

cool now we can ethicaly replace and devalue artists it waaas always about the database right HAHAHAHAHAAAAHHAHAHAHA

Seeing how in Midjourney's images the coca cola logo shows up, makes me feel resentment. I hope Adobe, other creators and new laws will lead AI art in a healthy and moral direction!

7:10 that's Robert DeNiro on the top right corner.

I only like ai as a way to help form ideas. for example for me, environments or character poses. but when people actually use ai and claim it as their own, i do not agree with.

AI is a good search tool

Wait. Are you telling me they actually did it the right way?!

Oh good new. Yes, unethical training has been the main thing keeping me from AI image generators.

I’m really brainstorming how I’ll be able to use these toys as research and reference purposes - I’m thinking that for the kind of art that I produce, that might be a big ask This definitely helps me get a better grasp on which apps might succeed more based on my needs, very interesting

I’m glad to see MidJourney getting some competition. Legally safe with an intuitive interface and the right price will destroy MidJourney. Does Firefly try to restrict content? MidJourney has a lot of words blocked and will ban you if you try to circumvent. Basically, anything the CEO thinks is bad, like “blood”.

This is so sick

Adobe firefly is rubbish

firefly has the cure