A.I Art in covers

[niobium]

How come A.I art generators are discouraged ? I know everyone wants public domain images for use on the covers, but Im sure making and using your own images and would not breach any legal issue

[TriciaG]

In short, copyright on them is murky. Most of the reasons given are here: viewtopic.php?t=98823

It was also mentioned that there have been instances where the AI doesn't modify the artwork it uses to generate a new image, so it can use copyrighted images and reuse them without permission.

Basically, until the US government clarifies the laws regarding AI and copyright, we don't want to use them, just like texts that are unclear in their copyright status.

[mightyfelix]

On a more philosophical level, the point was raised in the admin discussion that AI in general runs counter to the kinds of values that have always been celebrated at LibriVox, that is, sharing our human talents and ingenuity, spending labor out of love and passion. As people who celebrate that, it doesn't make sense to let computers do that labor for us.

But if philosophy isn't your thing, then yeah, the murky copyright status is reason enough.

[Alister]

Even if you don't use A.I. imagery as cover art, it may still be useful to you.

Since they essentially pull from a massive massive massive repository of imagery filtered by specific set of ideas or key phrases and then puke out thoughtless frankenstein blends, it is almost like tapping into humanity's subconscious psyche. Filtering your requests in something like Bing Image Generator with qualifiers like "Art Nouveau Book Cover", "Art Deco Record Sleeve" or "1970s Movie Poster" can be an excellent way to streamline the creative process when it comes to sourcing visual references. Something which many artists do anyway as part of the creative process (and I think concept artists do almost religiously). Of course, a google image search is just as useful

Edit ~

Here is an AI generated image I liked the look of, and alongside it my draft cover for a book I plan to narrate. . . eventually.

Hyperlink

It isn't my finest work, but it probably isn't my worst either.

[niobium]

Even if you don't use A.I. imagery as cover art, it may still be useful to you.

Since they essentially pull from a massive massive massive repository of imagery filtered by specific set of ideas or key phrases and then puke out thoughtless frankenstein blends, it is almost like tapping into humanity's subconscious psyche. Filtering your requests in something like Bing Image Generator with qualifiers like "Art Nouveau Book Cover", "Art Deco Record Sleeve" or "1970s Movie Poster" can be an excellent way to streamline the creative process when it comes to sourcing visual references. Something which many artists do anyway as part of the creative process (and I think concept artists do almost religiously). Of course, a google image search is just as useful

Edit ~

Here is an AI generated image I liked the look of, and alongside it my draft cover for a book I plan to narrate. . . eventually.

Hyperlink

It isn't my finest work, but it probably isn't my worst either.

but how come you get to use your own A.I image, but I cant?

[Alister]

but how come you get to use your own A.I image, but I cant?

Because laws. Unfortunately. It places limits on us. Both laws that have been written, and laws which Pollies will get around to writing. . . eventually. . . once they take a break from grandstanding and shouting about pointless things and do their jobs. But you see. . .

I'm not going to be using an A.I. image . Look closer.

Everything in my image on the right is done from scratch using assets which are public domain, or my own work. There are no assets in common. I'm simply using the image for inspiration - as a starting point to imitate. Imitation, or 'appropriation', as they call it in art, is always easier than working from nothing. Just don't use the image itself.

For good measure you should probably incorporate some fresh elements and add your own flourish and so long as there is no trademarks or copyrighted subject matter involved like a popular superhero, you should be golden.

Made using Inkscape and Paint.net. I recommend them strongly; they are free fantastic tools if you want to make covers.

Edit-

Here is an example in practice I found on Audible;

https://www.audible.com.au/pd/The-House-on-the-Borderland-Audiobook/B00FO542GO?qid=1689853470&sr=1-3&ref=a_search_c3_lProduct_1_3&pf_rd_p=771c6463-05d7-4981-9b47-920dc34a70f1&pf_rd_r=Z9FRDDDC2EV155AVSCZX&pageLoadId=Gjt4Rils3rrydZhZ&creativeId=adcc4fec-4d90-49d1-997e-8be21d68ce7f

From;

https://encrypted-tbn0.gstatic.com/images?q=tbn:ANd9GcRuuF4q3MpSdS-mDS3BsVu1-ZKG0DKisWc5zg&usqp=CAU

The text is out of copyright, but the cover art of that edition probably isn't. No worries; They've made their own. . . in the same style. It is their completely their own work though so they have the copyright.

[annise]

The advantage of using a google image as a base idea is that you have usually some idea of it's source but with AI you don't -and whether that is legal or not is not known .
It seems to me if I looked at a painting made by a 1940's artist and used my canvas and my paints to recreate it, I'd risk being charged

Anne

[Alister]

The advantage of using a google image as a base idea is that you have usually some idea of it's source but with AI you don't -and whether that is legal or not is not known .
It seems to me if I looked at a painting made by a 1940's artist and used my canvas and my paints to recreate it, I'd risk being charged

Anne

. . .In theory, I suppose it is possible. Unlikely, but popular artists aren't strangers to lawsuits. So it could happen if you did.

The originality of the image corresponds to the degree of copywrite protection. So if said painting is a portrait of Agatha Christie or a study of a Bowl of Petunias is liable to make a great difference.

That is however, if you are copying something wholesale. Nothing wrong with reworking it or drawing out design elements though. The trick is 'transformative' use. Changing the meaning or changing the aesthetic, is the ticket it seems. So for example, although my drawing upon design elements from the AI image probably doesn't cross the threshold of derivativeness where I'd need to justify, it does very clearly change the meaning of the image. In the House on the Borderland example, the aesthetic differs in that it is vector art; if they'd been capable of or driven to imitating the original more faithfully. . . That would potentially be more legally fraught.

As for the AI. Nothing I've seen suggests anything it spits out is 'undigested' enough that copying a copy is likely to put you in trouble. In a lot of ways the AI generators themselves seem to rest on the principle of 'transformativeness' when they draw from so much copywrited content all across the internet which is probably why they haven't been sued into oblivion. The specifically rely on the fact that what they produce is unrecognizable. Hard to sue when you can't even point to what has been copied! And even harder to not get thrown out of court.

They are essentially reducing human artistic expression down to the atomic level and piecing it back together, not just ripping out a face here and a car there and a visual element here. That is why complex elements like faces and hands and ships and military vehicles, among other things no doubt, often look wrong, warped or just plain wacky.

Of course, the law moves with all the haste of glacier. . . But we may see significant shifts.