Anyone using AI art?

[Aosta]

Well I made a quick mock up of Bjork's new album cover

[AdamMon89]

Well I made a quick mock up of Bjork's new album cover

Haha, very good, love Bjork.

[Quarmat]

Yes, with Midjouney. The whole thing blew me away. I work in graphic design and I am in total awe for the possibilities that AI is unleashing in terms of human creativity. I already used some AI generated visuals for my daily work and the collaboration between man and machine had never been so exciting.

Here's a selection of my early stuff.

[Fotu]

I just made this video for an original song using DALL-E for still image generation. It's DIY-ish, but nevertheless demonstrates how AI art will be useful for people like me, inept at graphic art, to express ideas visually.

[moofi]

01.jpg (227.65 KiB) Viewed 14615 times

03.jpg (153.63 KiB) Viewed 14615 times

04.jpg (191.29 KiB) Viewed 14615 times

[wendylou]

Created via text input using Stable Diffusion at DreamStudio Lite https://beta.dreamstudio.ai/dream

1201981133_a_hybrid_between_a_moose_and_white_goose.png (450.88 KiB) Viewed 14607 times

2865148705_andromeda_galaxy_inside_a_clear_glass_wine_bottle_on_wooden_table_in_kitchen.png (350.43 KiB) Viewed 14607 times

1585878712_spaceship_painting_in_the_style_of_john_berkey_.png (481.98 KiB) Viewed 14607 times

[wendylou]

2888482023_spaceship_painting_in_the_style_of_john_berkey_-3.jpg (341.37 KiB) Viewed 14605 times

[moofi]

Before going deeper into AI matter, I simply wanted to share this link. It´s Freemium and while there are some limitations regarding the AR e.g., with the free version you can already generate as many images as you would like to. Btw it´s based on pure Stable Diffusion. With the free version you are tied to styles (though some really good ones), then "realistic" style is pretty close to the pure Engine. One of the biggest helps is to include artists in your prompts like: "Magical tree, Greg Rutkowski, da Vinci".

http://wombo.art

[moofi]

it´s me.jpg (200.19 KiB) Viewed 14585 times

[bxbrkrz]

When you include the names of the original artists the machine learning phase was trained on, is that information always public, can it be obfuscated?

[tomusurp]

I haven’t used Craiyon but on midjourney for example, as stated in terms of use, as a free user you are allowed to use any image in their community under the Creative Commons non commercial license, meaning it can’t be for monetary gain.

But if you are using cover art for music you still need at least an active basic membership which also allows you a certain amount of your own generated images.

And in terms of owning copyright to any images you generate, you are allowed to claim copyright if you have a private feature on your membership which costs extra. This allows you to store your generated images privately and then if you post them somewhere on internet and someone uses it for substantial monetary gain, I mean you could file something like a dmca claim I think but still chances of that happening are slim.

Personally I just found out about this stuff and signed up to the basic membership. The thing is I use these arts for backgrounds in my photoshop where I always layer a picture of myself into it along with my name and etc. so I don’t mind if it is reposted or reused.

The issue isn't that you can't use the AI image for your album cover art, or even whether you wish to give away or sell your music with the artwork attached, you can do all of this. The issue is that you may not be able to claim copyright for that image yourself going forward as it may be deemed to be public domain. If that isn't an issue for you which it sounds like it isn't then there isn't a problem.
You just need to make sure that you understand the potential consequences and are comfortable with them.

From what I can gather, and again I say that this is only my opinion so please do your own research, it's no different than using any other image in the public domain. You do so in the knowledge that you don't own copyright for that image, and just as you were free to use it for your album cover, then someone else may also be free to use it for their own purposes.

A quick look at the terms of use for craiyon suggests that as long as you've not got an annual turnover of $1m then you can use images created commercially.
The only caveats seem to be that firstly you need to attribute the images used to craiyon (usually this is simply text placed on the cover, saying "image created using craiyon.ai" or similar (sometimes the company will require wording to be more specific, you would need to check).
Secondly if you plan on selling the image as an 'nft' (which you don't as far as I aware) then you need to pay them 20% of the revenue, and thirdly, that you are not guaranteed access to the site (they can terminate your access at any time and you have no legal redress, e.g. you cannot sue them for loss of earnings as a result). Again that's just my interpretation of their terms of service, but please check it yourself and get advice before proceeding.

I don't know if I've explained it without making things more confusing, but hope it helps!

[moofi]

With Wombo it´s commercially usable with rights from the get go, aswell with the free version. Then if you use it commercially there´s an attribution to wombo of some sorts required.
Besides the free usage it´s as well way more affordable compared to Midjourney. Bascially 5,49Euros/month with greater calculation speed constantly (around 13 secs for a 1024x1024 picture), no picture limit, generation of a 2x2 preview grid to select from via privately DMing a Discordbot, etc. There´s even a lifetime offer for something like 112 Euros. That´s just a little more compared to 2 months of MJ if you go for the option to generate in a private Discord channel Also DMing the Bot is private compared to MJ where MJ scans the private channel anyway. A pretty big plus in that regard is Wombo allowing adult content creation for Premium Users via the Discord bot as well. I´m quite a bit into generating gory zombies, something not allowed so far (and from what I´ve heard from the developers broadcast, MJ won´t allow adult content at all anyway). Currently MJ is still cream of the crops, then, from what I´ve heard, it´s rewriting the prompt and adding a style internally for their typical look. I´m not a hundred percent sure what AI they use specifically, then at least the saying is MJ using Stable Diffusion under the hood since the new beta as well. If that is actually true, you would be able to reproduce MJ´s style with Stable Diffusion anyway or at least get pretty close to it if you know how.

For anyone interested, here is Wombo´s Discord server to check out: https://discord.gg/wombo

I haven’t used Craiyon but on midjourney for example, as stated in terms of use, as a free user you are allowed to use any image in their community under the Creative Commons non commercial license, meaning it can’t be for monetary gain.

But if you are using cover art for music you still need at least an active basic membership which also allows you a certain amount of your own generated images.

And in terms of owning copyright to any images you generate, you are allowed to claim copyright if you have a private feature on your membership which costs extra. This allows you to store your generated images privately and then if you post them somewhere on internet and someone uses it for substantial monetary gain, I mean you could file something like a dmca claim I think but still chances of that happening are slim.

Personally I just found out about this stuff and signed up to the basic membership. The thing is I use these arts for backgrounds in my photoshop where I always layer a picture of myself into it along with my name and etc. so I don’t mind if it is reposted or reused.

The issue isn't that you can't use the AI image for your album cover art, or even whether you wish to give away or sell your music with the artwork attached, you can do all of this. The issue is that you may not be able to claim copyright for that image yourself going forward as it may be deemed to be public domain. If that isn't an issue for you which it sounds like it isn't then there isn't a problem.
You just need to make sure that you understand the potential consequences and are comfortable with them.

From what I can gather, and again I say that this is only my opinion so please do your own research, it's no different than using any other image in the public domain. You do so in the knowledge that you don't own copyright for that image, and just as you were free to use it for your album cover, then someone else may also be free to use it for their own purposes.

A quick look at the terms of use for craiyon suggests that as long as you've not got an annual turnover of $1m then you can use images created commercially.
The only caveats seem to be that firstly you need to attribute the images used to craiyon (usually this is simply text placed on the cover, saying "image created using craiyon.ai" or similar (sometimes the company will require wording to be more specific, you would need to check).
Secondly if you plan on selling the image as an 'nft' (which you don't as far as I aware) then you need to pay them 20% of the revenue, and thirdly, that you are not guaranteed access to the site (they can terminate your access at any time and you have no legal redress, e.g. you cannot sue them for loss of earnings as a result). Again that's just my interpretation of their terms of service, but please check it yourself and get advice before proceeding.

I don't know if I've explained it without making things more confusing, but hope it helps!

[moofi]

double post, sry.

[moofi]

At least with wombo, using the app/web version or the discord bot would make it only visible to you if you are talking about the actual prompt. If you are referring to the actual image, it sometimes can contain a "fantasy"-signature, you cannot really decipher anyway from what I´ve seen. Like here:

.

karate cat.jpg (122.01 KiB) Viewed 14569 times

When you include the names of the original artists the machine learning phase was trained on, is that information always public, can it be obfuscated?

[moofi]

In addition, only if you generate publicly via Discord people will be able to see the actual prompt you used along (including the artists you included). Saying when using the app/web app or the premium Discord bot only you can see the original prompt. With Midjourney, this is possible aswell, though you would have to pay 50$/month for creating in a "private" discord channel.

[crimsonwarlock]

I have Stable Diffusion installed locally on my laptop. Doesn't cost anything, unlimited use and as private as it can be. This laptop doesn't even have a GPU, but this installation can also run on the CPU. It takes a while to generate an image, but it works.

[moofi]

Seriously though, running Stable Diffusion locally is pretty neat, especially if you have an according videocard. (From what I´ve heard you can merely generate at a lower resolution base in case you have less than 10GB video RAM). Then being a heavy generator, it would occupy the PC constantly while via a service it does neither take up ressources, nor cost the electricity. And those mentioned 5+ Euros are a pretty fair trade in that regard, let alone the extras like 2x2 preview grid generation. When having a proper videocard I´m gonna use it locally aswell. Especially for stuff like sampler-models, identical seed generation etc. Just wanted to share wombo because it feels like a pretty good alternative for given reasons. I created like 20k images during one month since I learned about AI-Art, something made possible via the premium bot use. Like said, no extra cost, nor hostaging the videocard either.

YOOOOOOO.jpg (141.56 KiB) Viewed 14561 times

I have Stable Diffusion installed locally on my laptop. Doesn't cost anything, unlimited use and as private as it can be. This laptop doesn't even have a GPU, but this installation can also run on the CPU. It takes a while to generate an image, but it works.

[crimsonwarlock]

Seriously though, running Stable Diffusion locally is pretty neat, especially if you have an according videocard. (From what I´ve heard you can merely generate at a lower resolution base in case you have less than 10GB video RAM). Then being a heavy generator, it would occupy the PC constantly...

As I said, I'm running it (currently on my laptop) on the CPU. I can do other work on the laptop while it is running in the background. I just have to make sure it gets enough memory to work in.

warrior_princess.png (555.84 KiB) Viewed 14559 times

[moofi]

How long does one picture take?

Seriously though, running Stable Diffusion locally is pretty neat, especially if you have an according videocard. (From what I´ve heard you can merely generate at a lower resolution base in case you have less than 10GB video RAM). Then being a heavy generator, it would occupy the PC constantly...

As I said, I'm running it (currently on my laptop) on the CPU. I can do other work on the laptop while it is running in the background. I just have to make sure it gets enough memory to work in.


warrior_princess.png

[bxbrkrz]

At least with wombo, using the app/web version or the discord bot would make it only visible to you if you are talking about the actual prompt. If you are referring to the actual image, it sometimes can contain a "fantasy"-signature, you cannot really decipher anyway from what I´ve seen. Like here:

.karate cat.jpg

When you include the names of the original artists the machine learning phase was trained on, is that information always public, can it be obfuscated?

[crimsonwarlock]

How long does one picture take?

On my laptop, which is very under powered, about 40 minutes. I'm going to install it on my studio machine today, to see how it performs there.

EDIT: have it running on my DAW now. It has a GPU, but not with enough VRAM. Running on my DAW, using the CPU, has cut down render time to 12 minutes.

[moofi]

Yeah, well, currently I´m creating up to 800 images a day, saying 12 minutes rendertime would be a little slow here
With an RTX 3080 you can get one image in around 20 seconds. Even then, though quite acceptable, with wombo I can create a preview grid of 4 pictures within 13-20 secs and only choose the ones I like to render big while the upscaling from a chosen one takes like 1-2 seconds. Then wombo isn´t aiming for seeding, saying it will always be a new outcome generated from the prompt. The seeding plus using varying sampler models is something I´m going for locally eventually when I can afford a new videocard.

How long does one picture take?

On my laptop, which is very under powered, about 40 minutes. I'm going to install it on my studio machine today, to see how it performs there.

EDIT: have it running on my DAW now. It has a GPU, but not with enough VRAM. Running on my DAW, using the CPU, has cut down render time to 12 minutes.

[TritoneAddiction]

Glad this thread took off. I have a lot to explore. I'll use this thread as something to go back to over time to check out different options.

[crimsonwarlock]

Yeah, well, currently I´m creating up to 800 images a day, saying 12 minutes rendertime would be a little slow here
With an RTX 3080 you can get one image in around 20 seconds. Even then, though quite acceptable, with wombo I can create a preview grid of 4 pictures within 13-20 secs and only choose the ones I like to render big while the upscaling from a chosen one takes like 1-2 seconds. Then wombo isn´t aiming for seeding, saying it will always be a new outcome generated from the prompt. The seeding plus using varying sampler models is something I´m going for locally eventually when I can afford a new videocard.

Running locally indeed gives more options. As I can use the same seed to re-render the image with little tweaks, generating hundreds of images a day is no real use to me. I like the option to use this to actually develop an idea instead of just randomly generate stuff. For what I (try to) do with it, 12 minutes for a render is acceptable. I used to run DKB-trace on MS-DOS back then (yep, I'm old) and rendering a 320x200 image (the ones with all the shiny spheres) took about 72 hours

[moofi]

04.jpg (151.53 KiB) Viewed 14534 times

02.jpg (166.16 KiB) Viewed 14534 times

01.jpg (151.38 KiB) Viewed 14534 times