You forgot: play possum. Deceive until a person shows a second nature. Could fall into 2, but necessary for gain or advantage but for evil actions.
i’m afraid to say those are great! haha…
mediocre at best! small smile.
seriously, tho, i like that they’re horizontal flowing…
Thank you so much! 
Actually the flow was never specified. The prompt asked to seek the light.
The aspect ratio, it’s just because I love to see this full screen on my monitors. 
ah even better! my first thought was a tree growing up in that style but the side to side works better i think?
here’s one of my old dalle2ex styles i miss… any help recreating it? i don’t think i have prompts, but i possibly have more samples…
was a style for Clippy Quest and Long Live LitRPG… two idle / clicker games i made ages ago…
These are really cool. Mind if I try to replicate them?
Feel free to share here in thread for all. 
I’ve got no talent, I just string words together!
Seriously, tho, that’s close. A little too “finished” or “perfected” tho?
A lot of wood cut synonyms so cross hatch, ultrafine crosshatching, etc.
The other “trick” was to get back / fore -ground as B&W or color separately…
ETA: Sharing here from the DM for benefit of the community I love!
Summary
semantic_intent: | Create a square, high-energy action illustration of a fully armored medieval knight in the middle of combat. The knight should dominate the frame in an aggressive, dynamic stance, with a sword held horizontally across the composition as if attacking, parrying, or driving into an opponent. The armor should be full plate with a closed helmet and strong metallic segmentation, while the torso includes a bold black-and-white checkered surcoat or tunic layer that adds heraldic contrast.
The scene should feel explosive and immediate rather than calm or ceremonial. Surround the knight with kinetic bursts of black ink lines, radial slash marks, paint splatters, and energetic color eruptions, as though force and impact are tearing through the page. Include a partial secondary figure at the lower right edge or foreground to imply an active confrontation, but keep the main emphasis on the knight.
Render the image in a mixed-media style combining heavy black ink contouring, scratchy hatching, etched graphic marks, and loose watercolor or ink-wash splashes on textured paper. Use a vivid palette centered on black, white, steel gray, hot magenta, electric cyan, yellow, red, and orange accents. Emphasize graphic-novel intensity, tactile hand-made texture, and medieval combat drama over photorealistic depiction.
constraint_control: |
Preserve:
- square composition
- one dominant central armored knight
- horizontal sword as a strong compositional anchor
- full plate armor with reflective steel-gray highlights
- black-and-white checkered torso garment
- aggressive battle pose with bodily torque
- explosive background made of ink streaks, splatter, and radial force
- vivid accent colors: magenta, cyan, yellow, red
- mixed ink-and-watercolor tactile surface
- graphic-novel / fantasy action energy
Avoid:
- static posed character portrait
- clean polished 3D fantasy render
- soft airbrushed digital painting
- historically rigid realism with muted palette only
- overly detailed background setting
- ornate decorative court scene
- excessive gore
- weak or passive stance
- losing the central readability of the knight silhouette
i think your “mixed media” is key likely. thanks!
I’ll share the optimized version once I get it done right. Might take a full day. There are some things I need to tackle to make it user friendly. Check the DM and you’ll see why. That’s WAY too long.
@DysTopia was a lot closer to the style I was trying to replicate… i should’ve saved some of the prompts… wasn’t as contradictory as i might’ve hinted… but the words selected and order affected the output a lot more… that was back with GAN/Diffusion not the new multi-modal models.
Even @DysTopia’s second one is a bit too overbaked and not raw enough? could be watermarking too maybe.
Anyone think up a new theme for July yet?!?
ooh you got me wondering if you could give it a real cross-stitch (?) pattern and ask it to “complete it” … or upload an image and get the pattern out…
Second one was an attempt to do an all in one prompt. Got too convoluted and lots of verbosity. 
wonder where it got cupcake? lol then after it realized ‘oh, it’s a heart!’ … kinda wild… this was for the heart pattern…
Creating finished cupcake cross-stitch art image
The pattern is a cupcake with frosting. I’ll render it as a completed cross-stitch on off-white Aida cloth, with no grid lines or legend, just the stitched motif. To make it look like a finished piece, I’ll include the fabric and maybe a soft tabletop background.
Fixing cross-stitch image shape to match pattern
The generated image turned out looking like a heart instead of the intended cupcake or ice cream shape. I’ll regenerate the image, emphasizing the exact silhouette—a double-scoop ice cream cone or cupcake-top shape, with two rounded pink scoops, a purple rectangular band, green trapezoid body, and a yellow cone. I’ll use the original pattern as a reference again.
Yeah, that’s one of the things i liked about dalle2ex… it was a small-ish prompt if i recall correctly… under 100 words for sure… but it would “improvise” (not really) when trying to fulfill weird prompts… that said, it could also go off the rails easily, so they’ve kept improving until ‘ordinary language’ words a lot better…
You mean the chances you receive styles that are congruent instead of dissonant? Or that are what you were already thinking about or working on? That would be pretty hard to answer. We can (ask the AI to) do math.
The odds of pressing the button once and getting that exact two-style output, in that exact order, are 1 in 229,440.
Because there is a 1-in-4 initial branch of the number of choices to include, you’re asking about replication of least difficult text.
If instead, you were asking about the chance of a 5-item being replicated, it would be about one-in-19 billion.