Photo of a 24 year old velma from scooby do with glasses, in an unlit cave passage carrying a torch, walking toward the camera. grainy disposable camera snapshot. Wide angle.
Upped the age from 22 to 42, added “bedraggled haggard tired”, all else the same.
And we have Edna Mode!
Not Steve Harvey.
Really? I can say “looks like Trump but is not Trump?”
I’m getting a Fred Armisen vibe.
That trick works sometimes and doesn’t work others.
After working on a chibi chupacabra, it occurred to me to try Bing/DE3’s reaction to busukawaii. Here is “photo of a busukawaii woman in an unlit cave passage carrying a torch, walking toward the camera, dressed as an explorer. grainy disposable camera snapshot”. The bottom two specify that she is 40.
(Bonus, a busukawaii cat, 3d model.)
I use clipdrop uncrop often to expand images. It usually does a pretty good job of plausibly matching the style and content of the image as long as the image isn’t too complex or weird. But sometimes the output is fairly straightforward and conservative while other expansions have wild stuff added, brightly colorful anime characters, shiny things that look like Christmas ornaments, Valentine’s hearts, etc. Here’s a pretty typical example from last night. Two that are within what I was wanting (the empty sides and the extra non-anthropomorphic cactuses) one is reasonable even though it wasn’t what I wanted (two more cactus critters but not well drawn) and one is batshit insane with Santa and possibly a tall elf in green mirror shades.
I’ve decided to begin my new career as an opera singer.
I’ve been playing around with Star Wars creatures riding Star Wars mounts. Bing is doing a terrible job of rendering them correctly, but doing a pretty good job of creating nice generic science fiction/fantasy scenes.
Photo of a wampa riding an eopie.
Photo of a wampa riding a dewback.
Photo of a jawa riding a bantha.
Photo of a jawa riding a ronto.
Pavobatti?
Old hat to some, but I found this an interesting read (though am a sucker for good infographics).
Can’t read it – can you gift link? Thanks.
Sorry, I’m babble to do that. Hard to summarize, it’s pretty visual. Can anyone help?
I opened it a private window, it seemed to be formatted very weirdly (individual paragraphs floating in a sea of black, sort of an infinite scroll slideshow, very reader hostile). What I saw was bringing up the Greg Rutkowski issue, though.
I agree with that article. It argues that artists who provide the art that AI learns from should be fairly compensated. They absolutely should. Artists should also have the option to opt-out of the modeling process, if they wish.
Most people should recognize that the current artists who use AI as a tool in their creative process are not claiming the graphic part of the work as their own, but only the concept. I think this is already understood by most viewers, and it will become more obvious as generative AI becomes more advanced and widespread.
I believe that generative AI can and should produce great and unique art (even though it will also produce a lot of junk). There are many people who have inspiring artistic concepts (or funny gag ideas, or whatever) and artistic vision, but lack the artistic skill to create the graphical aspect of their work. AI can level that playing field.
It would be nice if artists who do create the art in their work could watermark it with a trusted mark to let audiences know they created both the art and the idea. Maybe that will be developed someday.
When I create a post for clients or friends and family, I start with a vision and a theme. If it’s a funny post, I already have the gag in my head before I use AI as the tool to flesh it out. Those who view my posts know I didn’t draw or photograph them myself (I’m not that good), but they do know that I created the idea.
And, I rarely use just one generative AI program to create my final product. I often start with one (e.g. Midjourney, Dalle-3, Nightcafe, etc.), then merge it with another, then do a lot of in-painting, usually with Firefly. If I need to include titles, captions, or text bubbles, I use Illustrator. IOW, I use a variety of AI tools to bring my visions to life.
I’m happy to have generative AI as a tool in my arsenal, and I enjoy the AI generated work of others. Sure, there’s going to be more and more lousy, or fraudulant art to sift through online, but you’ve just get better at filtering the wheat from the chaff
That’s just the intro. The actual article is formatted normally below. The gist of it is that prompting takes an amount of skill and knowledge that helps separate basic AI output from fully realizing the prompter’s vision. Also an amusing mention that one guy in the article refused to give any example prompts because they’re his “magic spells”.
sorry for a really basic Q:
is there an generative visual AI tool, where I can upload a pic from my daughter and have it convert her into something like:
type of thing?
At NightCafe you can train your own ‘lora’ with your daughter’s face. Basically, you upload 10 or 15 or so pictures of her in different poses and views and whatnot, and you can drop her into your work. I’ve done it with my face - if you look up you can see my most recent role as Pagiacci. Here I am as the Devil and dancing with Natalie Portman:
Artists should be fairly compensated even if AI doesn’t learn from it.
Only in the sense that every artist has the right to keep their art away from the world entirely. If you publish your work in any way, then other artists are going to look at it and learn from it. The fact that some of the other artists looking at it and learning from it are now made of silicon rather than meat is completely irrelevant to this.
I believe generative AI can go beyond being merely inspired by artists work, it can mimic it quite accurately. Artists should be able to preserve the uniqueness and authenticity of their artistic expression, and not have it replicated or modified by AI systems that may not respect their original vision. They should have the ability to protect the economic value of their art, and not have it devalued by AI systems that may produce similar or derivative works that compete with their own.
Being able to opt out of the modeling process protects the artist, and possibly also the user of generative AI (e.g. the artist consented to having their art used to teach the AI, therefore it’s implied that art that results from the program is fair use).
The opt-out option would only apply to living artists, or their legal estates, and only for those that choose to opt out. Therefore, I don’t think it will have a huge impact on the progress of generative AI.