but it destroys only what it considers to be shackles limiting our vision.”
― Salvador Dali
The seed image was initially created on DALL-E2 in Nightcafe. I cooked it using the same prompt on Deep Dream Text 2 Dream in Artistic mode, at 50%, and evolved HD 10% plus additional Photoshop work.
MONSTER MASH - CTHULHU'S LAIRinitially created on Nightcafe and evolved on Deep Dream Text 2 Dream
NIGHTCAFE CREATIONS
I got an insane amount of free credits on Nightcafe, so I use them:
I have been quite involved lately creating AI images on Nightcafe. Recently I joined a chat group of creators that belive in sharing prompts. Already prior to this I had cooperated with a creator/artist who had also not hidden his prompts.
The amazing thing is that with the same prompt you could get many different results as well as styles.
"The surrealistic monster landscape of H.R. Giger - assisted by Arcimboldo and Bosch, highly detailed extremely detailed fantasy oil on canvas imperial colors fantastic view hyperrealistic cinematic postprocessing Hieronymus Bosch Giuseppe Arcimboldo H.R. Giger Surrealism Giger Alien, fantasy illustration, detailed painting, deep color"
Negative Prompt (set to automatic)
"ugly, tiling, poorly drawn hands, poorly drawn feet, poorly drawn face, out of frame, extra limbs, disfigured, deformed, body out of frame, blurry, bad anatomy, blurred, watermark, grainy, signature, cut off, draft"
I had used this prompt across platforms, sometimes tweeking it, and with additional moderations. Occasionally I mix it up between Deep Dream Text 2 Dream, DreamWombo and Nightcafe using respective results on those other platforms and sometimes switching back and forth using results as seed images. AI can create thousands of different iterations, so 'evolving' one result can give quite a few interesting additional results. For the most time I opt to create four images at medium resolution on Nightcafe, the odd time nine. The best ones I 'upscale' and also publish separately.
Here the most recent examples on Nightcafe created with the brand new (Stable Diffusion) SDXL mode:
MONSTERS EVERYWHERE
SURREALISM ON STEROIDS - EVOLVING
This information below is copied over from my previous post - in the meantime, a Stable Diffusion XL mode has been added, which is also accessible via the drop-down menu
Here a brief explanation how things work on Nightcafe
You have a choice of 5 Algorithm:Stable Diffusion - DALL-E2 - COHERENT CLIP guided Diffusion - Artistic VOGAN+CLIP - Style Transfer
The first two are on the new advanced creation page, the others on the old style page. But with the VOGAN+CLIP on the old style page you can also create videos, but it costs quite a few point to do so.
It is a bit tricky to get to all the choices, but briefly you need to look for links to open those, like in Stable Diffusion for prompts you can open the Advanced Prompt Editor the simplest one are the Presets, then you have the Moderators and the one I go to all the time My Favorites (you can save any prompts from your, as well as any other creations that didn't hide them, in that file and reuse them.
On Settings look for All Settings to open more choices. This is important for a number of reasons: for one thing, you can now choose your output image size, the run time, the Aspect Ratio and many more, but one very important setting for Stable Diffusion would be the Model where you can choose between eight different models (I use mostly v1.5, but v2.1 gives you quite different looking images).
Most times under Prompts I leave the default Negative Prompt setting to On, but I had saved some negative prompts that had worked good for others in my Favorites.
You can also add second, third, and more different prompts and assign different weights to them.
Something else to tweak would be Clip Guidance and Sampling Method. Most times I leave them at default.
here are a couple of cards I created most recently
I could be milking these images individually for separate blogs but that is not me. I post maybe once or twice in a 2 week period and just hope you like my work, upvote and follow me.
Visit my website
▼
PRINTS OF MY ARTWORK AVAILABLE HERE:
Bonus Special: IF NOT NOW, THEN WHEN?
should you not be on the blockchain, or have no upvote power, or this post is already older than 7 days, you can always just buy me a coffee.
Was KI schon alles kann.... unglaublich 👻
ist aber nur künstliche Intelligenz, also es kann nicht selber denken. Es kommt aufs Training an, und da hat sich schon viel geändert seit den primitiven Anfängen. Ich habe auf Nightcafe noch immer ein paar solche gespeichert. Das Wichtige ist wie man Prompts schreibt und manipuliert, eben wie im Blog geschrieben.
KANARY IN A MINESHAFT
das war einer der ersten vor ein paar Jahren.
Noch kann sie es nicht...
Damn these really are on steroids :D I think I once again need to go back to wombo because those results you posted at the bottom look incredible, it must have had yet another upgrade since I last tried it !
there are a few styles that work better than others. The 'pretty' ones are from 'Flora' (there are two, Flora v2 and Flora). It hardly matters what you write for prompt, even the seed image. You get 'pretty' results that may be useful to mix in with other stuff or collage with in Photoshop. I get a kick out of mixing pretty Wombo results with my freaky type of prompts on Nightcafe.
Here is one more recent result that I had fun with after on Nightcafe, funny thing is, my prompt was about 'Bellmer Dolls' already, but on Wombo, it turned pretty: https://dream.ai/listing/0441ff00-5d2d-40ea-9999-d0f0dd8f75e3
My Profile on Wombo: https://dream.ai/profile/the_mystic
So I updated and tried it again, unfortunately a few of the styles that used to be free are now premium only but your right the flower one pretty much always gives really cool results!
I use them mostly for seed images on my other apps. But what I had not done yet is combine images in Photoshop, or clone stuff from Wombo into other digital stuff. btw, just wondering what the 'expert sleuths' would do with it if I cloned AI into one of my 'real' paintings; something I had been thinking about as well. It would be a digital collage. I had in past also cloned details of my paintings into actual photographs, and since those details were painted photo-realistic, it is impossible to tell which is which. (Like, the exif data would show the starter image data plus the dates of when it was worked on in photoshop, but not any data on the additional images).
Well I don't see anything wrong with that, I have been doing digital collage for ages and using AI for assets, much more diverse than using royalty free or other images on the Internet. This could be a new job though, AI sleuth 😂 I actually did see a guy get outed on twitter for using AI images and photoshopping them onto canvas and sketchbooks so it looks like he made it entirely, things are getting wild 👀
This melted my brain. Great imagery.
The rewards earned on this comment will go directly to the people sharing the post on Twitter as long as they are registered with @poshtoken. Sign up at https://hiveposh.com.