Computer Generated - Artwork made from text prompts!
Stable Diffusion is currently the best free and or paid public domain AI art generator out there!
Here's where to generate things for free using HuggingFace AI Community or Kaggle for absolutely free!
Alternatively you can find a Google Colab online of Stable Diffusion
https://huggingface.co/spaces/stabilityai/stable-diffusion --Long wait queues.
Instructions for Kaggle are in it's code. You get 37 hours a week as a free limit per week.
https://www.kaggle.com/code/scratchpad/notebook3f89987b34/edit --Fastest Option currently
The AI (artificial intelligence) DreamStudio uses Stable DIffusion AI which generates CC0 1.0 licensed images! it's a browser based website program.
License Note: https://stability.ai/blog/stable-diffusion-public-release
This service is at this site: https://beta.dreamstudio.ai/dream
Sweet! I've been hoping someone would finally make a CC compatible AI.
Unfortunate it's not a free service though.
Great find.
It does work now for free using Kaggle is the easiest way.
If you want to create a lot (by alot I mean hundreds or thousands) of prompts then I recommend Midjourney instead, their $30 a month plan gives you unlimited number of queries (in relax mode which is slower however) AND 15 fast GPU hours which roughly translate to 900 fast queries. Their ToS also gives you all rights for your images.
So far I'm making like 2000 images a month.
I would likely use Discord's Midjouney if I understood entirely how it worked. I am used to the simple interface personally. Also there's alot you have to read important info wise that is tough to catch up to. Does Midjourney allow more creativity in what you can generate/ I thought it had restrictions on like copyrighted material, certain depictions, and you can't make hilarious things like on Craiyon. Idk I could be wrong about that. I'm already in Midjounrey have been for awhile. Just never tried using it. I lack creativity to come up with anything for a prompt. Also I am more afraid of messing up with it.
Yeah it a nice value for money and furthermore, if you put it into a revenue stream of some sort that dose make money back, it is great but that is on you.
It does burn credit 'fast', but it someone's cpu/gpu being utilised, so no free option there.
p.s. works nice, so much potential.
sustainable_concepts._loop_epic._green_and_plue_wibe.png 1.3 Mb [3 download(s)]
Regarding restrictions on copyrighted material, certain depictions, etc: You may only use MidJourney output for personal use unless you are a paying member. In which case you're allowed to do whatever you want with the output provided it doesn't display any trademarked brands or depictions of real people's likenesses (which is actually pretty hard to get it to do accidentally)
--Medicine Storm
Yeah the discord interface is confusing, they said they're spending a ton on developing an actual web interface on their site. I don't know if MJ has more "creativity" than other services (you could check out Youtube comparison videos), but MJ does allow you to guide it in a particular direction very well. For example you want to make a very specific thing like the side view of a futuristic car, in pixel art style to use for your game, MJ first gives you stuff that you don't like, but after some variations you found something that resembles what you want, so you tell MJ to make variations of that specific sample, and repeat the process until MJ starts churning out an infinite number of cool looking and unique car sprites for your game (thats what I did). And because relax mode allow infinite queries, you can do this forever!
You don't need prompt crafting voodoo skills imo, I usually start with a very general prompt and then keep making variations of the samples that I like until MJ knows what I want and produce an infinite amount of it xD
MJ does ban many many keywords to prevent people from creating pornography, extreme gore, or some politically charged content (They say they don't want to be banned from China and they think that allowing Chinese users to make use of MJ for creativity outweights the ability to make Winnie the Pooh memes, take it what you will).
I'm wondering why OGA doesn't support the CC BY NC license?
Because it makes the assets effectively useless, unfortunately.
95% of all developers who come to OGA for the assets are using them in either an open-source project, or a revenue-generating project. -NC licenses are incompatible with both.
Here is an article about German courts conclusions regarding -NC licenses: https://www.techdirt.com/2014/03/27/german-court-says-creative-commons-n...
EDIT: here is the PDF document mentioned in the article: "Consequences, Risks, and side-effects of the license module Non-Commercial -- NC"
That article isn't what we've based the decision on, but it illustrates the problem well. Also, OGA is hosted in the US, so we typically don't go by German copyright law, but the results are basically the same everywhere in the Americas, Europe, and Japan: The meaning of "non-commercial" is very different from a legal standpoint than it is from a 'common understanding' standpoint. Even indirect generation of revenue by the free hosting site's sponsor may legally qualify as a "commercial use".
--Medicine Storm
I just added the free method of using the AI (Stable Diffusion/DreamStudio) it's a Google Colab notebook. Try it out the setup is seriously only the longest taking thing. Which is actually not even the worst waiting time. The generations for 4 images take about 60 seconds. You can also adjust the amount of images based on columns and rows.
Replaced the Google Colab with Kaggle. It's a little nerve wrecking making sure to shut off the virtual machine. However they give out a great portion a time for each week for use. WHile Google Colab will completely block you out for using their service freely for an entire day just by using it some time concurrentl. I believe it's better to be able to choose when to use your time for. Also I really messed up the Google Colab. So this rectifies that.
Artbreeder Collage added stable diffusion it's extremely fast however it works by image shapes for example.