So maybe you're familiar with the weird world of DALL·E, but how much do you know about #Midjourney, the OTHER super-duper text-to-image AI?
Well, between you and me, it's gotten REALLY GOOD recently – here are some stunning examples that show why you should give it a whirl 👇🧵
DISCLAIMER: all images have been plucked from the stunning community gallery that Midjourney offers paid users - of which more in a bit!
PART ONE: SHE HAS THE *RANGE*
@Midjourney perhaps has a reputation for creating amazing but very AI-y images like these. Super awesome, but with the v3 model, MJ is capable of so much more...
🎞PHOTOGRAPHY
In theory #midjourney doesn't really 'do photographs': but that's not totally true. It DOES generate super-sharp convincing images that, if not quite 'normal photos', are definitely 'photo realistic.'
📸 PROMPTING PHOTOGRAPHY
And yep, just like #DALLE, MidJourney understands all kinds of photography prompts for gear and setup, like fish-eye lenses, Kodachrome film stock, Polaroids or long exposures.
🧑🏾🎨 2D ART
Onto illustrations! #MidJourney has the chops to deliver everything from vector-style graphics and watercolor paintings to ballpoint sketches, plus styles from a host of famous artists like Van Gogh.
🗿 IMAGINARY 3D STUFF
Can Midjourney build it? Yes she can! From dollhouses to actual houses, mocking up 3D work is just a prompt away with MJ.
🤪 SILLY STUFF
Who can resist some cheeky engagement-bait?
Jesus taking a selfie, with a GoPro, on a surfboard? Yep, it's all possible here.
PART TWO: NO RULES*, NO MASTERS! *OK, FEWER RULES.
Want to make content about illness, war, suffering, politics + heavy topics on DALL·E? (Or celebrities, for that matter?)
Right now, that's all BANNED. But on easy-going #Midjourney, only real 'adult' stuff is blocked. Go wild!
PART THREE: SHE'S GOT HER OWN VISION
Good stuff so far, but real talk: Midjourney will take liberties with your prompt if she wants. e.g, compared DALL·E (L), MJ (R) couldn't quite stick to the rules of pixel art...
When it comes to a prompt for 'outsider art', #DALLE was quick to adopt a classic scribbly look, whereas Midjourney is more like: 'how about an artist with a bit of formal training?'
Midjourney also has some go-to aesthetic crowdpleasers - yes, your image looks amazing, but be warned: she might be sending very similar images to everybody else!
OTOH: I tried prompting for some general images around work/offices/mental health and got usable results straight off the bat, that all look like they could go well together. No finickity prompt engineering required. Seriously impressive!
PART FOUR: PRAISE BE, IT'S LANDSCAPE + PORTRAIT!
Landscape / portrait images are only a prompt away! No more unfortunate cropping or prompt-consuming stitching: just sweet panoramas, cinematic widescreens, poster layouts & full-body portraits in an aspect ratio of your choice.
PART FIVE: USING MIDJOURNEY.
Somewhat at odds with the overall calibre of the tech, MJ exists only as a Discord bot. IMO, this is partly why people are sleeping on it – it's a bit confusing to the Discord-unfamiliar squares like me.
It's not even Slack!
Unlike #DALLE2, all generations are saved *automatically* (no more lost work, yay!) while everybody's pics+prompts are public + searchable by default, so you can browse images and prompts all day long - THIS IS SO FREAKING USEFUL.
💸 At $10/month for 100-160 images, vs #DALLE's $15 for 115-460, the pricing is hard to compare directly, but I expect comes out pretty similar? $30/month gets you unlimited (but eventually 'slowed down') use.
It's $50/month if you want to keep all your pics + prompts private.
IMO as a visual 'product' Midjourney is easily the most market-ready: you could easily imagine MJ tied into @Pitch or @Canva, or a poster-printing service: anywhere users want to create good-looking graphics fast (and most of 'em won't want to 'learn prompt engineering' first)
Wanna know more? Check out the full explainer on @Midjourney here:
And for a weekly dive into the latest creative AI tools (and what they all ✨mean✨) sign up to my newsletter, PROMPT/RESPONSE: promptresponse.substack.com
If you've tried both, which do you prefer?
• • •
Missing some Tweet in this thread? You can try to
force a refresh
Today's #AIart prompt is... not a word, but pictures!
🎨 You can use #MidJourney to combine dissimilar source images to create surprising new effects – kind of a whole new take on the photocollage!
Here's how 🧵👇
Simply paste the URLs of two (or more!) images in the prompt – no additional text necessary – and MidJourney uses #AI to combine the concepts from both into a brand new image. Fun!
For some image combinations, the output is pretty predictable, as the concepts are not mutually exclusive.
For best results, choose source images with a similar aspect ratio (portrait/landscape) and choose an output ratio to match (e.g: --ar 3:2 for a landscape output)
Part of me does question how valuable a super-abundance of AI-driven content will REALLY be, even if it's really 'good.'
If there's one thing the world does not feel short of at present, it's 'content.' In fact, a lot of the time it feels like there's already Too Much!
With more humans, and easier publishing, humanity creates more and more each year, whether that's music or literature or film or whatever.
PLUS, it accumulates - the archive of all our previous creative output is now more readily accessible than ever. It's sort of stressful!
And that's just the 'classic' content formats - there is of course the relentless firehose of social media, from Twitter to TikTok, and Substack, and other content that 'flows' like the news.
With just a brief example you can teach #GPTchat how to generate #AIart prompt templates for any topic (and yes, you can get it to dream up the topics too!) and then create matching prompts
Then you can get it to generate random descriptions based on those templates...
Which text-to-image #AI can then turn into images!
(Asking #chatGPT to generate formatted prompts for random fashion shoots, then posting the results into #MidJourneyV4.)
HD photos in thread - see the prompts in alt text! 🧵👇
1/5 💖💛
2/5. The example on the right shows how creative #AI prefers coherency over surrealism: once it had decided he had a 'regal' expression, it made the background 'a palace.'