I still think you might need to do some basic editing to get the results you want.
Yeah. Much like real photography, or other crafts, most of the work is done in Photoshop (or the darkroom for film development).
Writing is rewriting.
Woodworking is 20% construction, 80% sanding.
It’s totally worth the effort.
Though I must say, I’m growing partial to prompts with “oil painting.” You get more flexibility with what’s realistic but it’s also easier to fix imperfections in editing. Smudge tool blends in perfectly. Not as easy to do with photographic style.
Bit the financial bullet, bought a Midjourney subscription, am now having much fun creating character drafts. (/relaxed mode is a must for draft work)
Hannah Gavish (23) and her Alpina B6 (2016 model)
^The Alpina should be black, but the green is winning me over, and what’s with the taxi sign on the roof?
Update: Decided to give the character Dov another attempt.
^A twenty-year-old Dov Shaul, proud member of the Australian Army Commandos.
Just discovered the Midjourney AI has some trouble creating portraits of twin characters, or I’m still having trouble finding the right prompts…To find the right answer, you need to find the right question ~ A quote from my Information Management studies…The prompt section also lacks a spell checker.
I attempted some experiments with vehicles and discovered Midjourney does not comprehend some military types, like the Merkava tank or the Hind gunship. However, it does okay with civilian vehicles.
A twin-engine jetliner over Tel-Aviv. I tried a Tu-154 jetliner (owned by Karen), but the AI did not understand the unique model (a Soviet era tri-engine design). So maybe Karen upgraded to something more efficient and luxurious.
^Need to remove the top row of windows. Nice of the AI to include the ECM / EW pod beneath the fuselage…for the privacy and security conscious passengers…
A metallic-blue RAM1500 beach scene
In a short story, Freyja buys a blue RAM1500 for Dov as a surprise gift so they can tour Australia’s northern beaches in relative comfort…Blue being Dov’s favourite colour…
This is really retro, almost could expect that hairstyle in MLK’s era…or further back.
Green Audi R8 and S5 parked on the driveway of a European villa
I don’t know if the five-ring emblem was a mistake or the software trying to avoid a copywrite issue.
Interesting work with the lighting and reflections.
It’s translating it as fake words instead of a logo, at the least. Some look close to real ones. 3rd: Feed, 4th:Cesar, 5th:Creed.
At first I thought that a picture of some actual twins would help the AI but Nikki and Gabby here is triggering my ancestral fight-or-flight mode with their plastic surgery faces.
There’s something wrong with this image and it scares me.
I’m reminded of a window sticker on a 1960s Chevy: Classic not plastic.
Inspired by a YouTube video about Japan’s car culture…
Prompt: RAM1500 Itasha wrap, photorealistic, ultra-detail, 4K, car show scene, DSLR
…and some RAM2500s…
I think these are closer to RAM3500s / Super Duties.
I tried prompting anime characters and J-Pop Itasha but they didn’t work. MidJourney couldn’t comprehend the Holden Sandman station wagon either (heresy). However, it has a passable understanding of the Holden HSV, so there is some hope.
All of your car images are awesome. Makes me want one like it so bad.
Built for a time fuel is / was cheaper than water…
I tried prompting for Bosozoku styled or 6x6 RAMs but I didn’t get anything close…maybe it’s too unique for the software…I did receive some interesting concept art though.
Once you start, there’s no end…
This is getting silly. Back to character portraits.
And the first lawsuit has rolled out.
Today, we’re taking another step toward making AI fair & ethical for everyone. On behalf of three wonderful artist plaintiffs—Sarah Andersen, Kelly McKernan, and Karla Ortiz—we’ve filed a class-action lawsuit against Stability AI, DeviantArt, and Midjourney for their use of Stable Diffusion, a 21st-century collage tool that remixes the copyrighted works of millions of artists whose work was used as training data.
We’ll see where that goes.
I read recently that those using Midjourney should not upload images of real people for prompt references, as that leaves the user exposed to legal issues. I was considering citing the names of famous people for my character portrait prompts, as others have done in the discord feed, but now that idea is binned…I’m getting more preferable results without them anyway…I had also considered citing the styles of my favourite photographers and anime artists (by name) out of curiosity, but that will wait too.
I wonder how much AI art falls under Fair Use or Creative License if it’s not used for profit / commercial means…and which country / region holds legal jurisdiction…
Good artists copy, great artists steal ~ Picasso
To steal from one source is plagiarism, to steal from many is research ~ Steven Wright
Fair use has always been widely debated. How come parodies are okay but fanfics are not? Because once upon a time a court made a decision that has become a law. It has nothing to do with ethics or quality of work (I’d argue that a parody is worse than a fanfic since it mocks the original while a fanfic celebrates the original). Could one pass off a fanfic as a parody and defend it with a “humor is subjective” clause?
So this lawsuit may be a starting point for all generative image laws going forward. It’s worth following.
There are two characters for which I used a real person in the prompt but I only used the results that ended NOT looking like that person. The appeal of that is that it’s easier to get a more realistic-looking face and to control the type of face. But yes, I wouldn’t want to use those for commercial purposes. Not until the legal guidelines are more defined.