Stylized portrait of a nun in black habit, face turned slightly upward in profile, oversized round sunglasses with ornate gold filigree frames and reflective magenta-pink gradient lenses, smoking cigarette with wisps of smoke curling upward catching neon light, massive glowing neon red-orange circular halo behind head casting dramatic rim light onto silhouette, heavy gold chain necklace with large textured gold cross pendant, skin rendered in iridescent teal and cyan tones with subtle purple shadows in hollows, detailed tattoos visible on hand and wrist, graffiti-covered concrete wall background in deep blues teals and magentas with abstract tags and layered symbols including teal Chinese character in lower right corner, dramatic rim lighting from neon halo creating glowing edge on habit, cyberpunk aesthetic with religious iconography subversion, digital painting with thick oil paint texture and visible brushstrokes, high contrast saturated colors with crushed blacks, cinematic atmosphere, moody and rebellious energy --ar 2:3 --style raw --s 750 --v 6.1
Click the prompt box above to select it, then press Ctrl+C (Cmd+C on Mac) to copy. This prompt deliberately mixes painterly and photographic instructions. Decide which look you want, and delete the words for the other.
This prompt contains an internal contradiction, and it is worth pointing at it before anything else, because it explains most of what makes this image work and most of what will fail if you change it.
It asks for digital painting with thick oil paint texture and visible brushstrokes and, in the same prompt, for hyper-realistic skin texture with subtle purple shadows and a photographic lighting setup. Those are two different media with two different behaviours. The prompt is holding both at once and leaning on the tension.
Why the contradiction is not fatal
An oil painting does not have photographic skin. It has brushwork that describes form, and form is what a portrait carries. A heavily rendered surface with visible strokes can read as more convincing than a smooth render of the same features, because the strokes describe how a human built up the image rather than pretending the image was captured.
That is the real argument for the painterly route in AI portrait work, and it is not a small one. Photorealistic faces are the single most-trained subject in the image models' training data, which means photorealism is where they are most confident and most exposed. digital painting with thick oil paint texture moves the request into a domain with a much larger, much more forgiving error margin.
The cost is that you lose the things photography gives you. Skin pores will not be pores, and any instruction that demands them will produce texture that fights the brushwork.
The rim light is the structural decision
massive glowing neon red-orange circular halo behind head casting dramatic rim light onto silhouette and, again later, dramatic rim lighting from neon halo creating glowing edge on habit
Rim lighting is a real and specific technique: light placed behind and slightly above the subject so that only the edges catch. The rest of the figure falls into darkness. It is the standard way to separate a dark subject from a dark background without lighting the face, and it is the correct choice here because the face is turned upward in profile and the light is behind the head.
The editorial prompting guide describes rim light as one of the setups worth naming explicitly, along with the specific instruction to give angle, height, source size and shadow behaviour rather than a mood word. A halo behind a head is a large, close, bright source — closer to a practical light in frame than to a studio strobe — and saying where it is relative to the head is what determines how wide the edge glow runs.
If the glow spreads halfway across the habit, the light is too close or too large. Narrowing it usually fixes it.
Iridescent skin is a colour decision, not a lighting one
skin rendered in iridescent teal and cyan tones with subtle purple shadows in hollows
The second half is the detail that makes it work. purple shadows in hollows specifies where the colour goes, and hollows — the eye sockets, under the cheekbone, the neck — are where a cool key light naturally falls off. Naming the shadow colour in those specific places gives the model a placement rule rather than a general tint.
Real skin shadow does not go grey; it shifts hue as light bounces off surrounding tissue. Describing that as purple in the hollows is a compressed way of asking for the same effect, and it is why the skin reads as lit rather than tinted.
The teal skin and the warm halo are the opposite ends of a temperature split, which is the same structure described elsewhere on this site as colour temperature warfare: two areas separated by hue as well as brightness, so the eye reads depth before it reads detail.
Gold filigree against magenta lenses
oversized round sunglasses with ornate gold filigree frames and reflective magenta-pink gradient lenses
Filigree is a genuinely hard material for this kind of model: very fine repeating curves, high contrast, and a tendency to dissolve into noise when the pattern is dense. On a small round frame it usually holds better than on a large rectangular one, because there is less of it.
The gradient on the lenses does the same job it did on the iridescent pair elsewhere in this set — a gradient implies angle-dependent colour rather than flat tint, which is what makes a lens read as glass.
Where it fails is predictable: if the filigree turns to mush, removing ornate and keeping fine usually helps, because the model is more willing to draw a fine pattern than an elaborate one.
The stylize value is the highest in this set, and here it is the right choice
--s 750, near the top of the working range.
Midjourney's stylize documentation gives the range as 0 to 1000 with a default of 100, describing high values as allowing the model to interpret freely and stray from the prompt's exact detail. The print-style guide splits that into working bands, with 300 to 750 identified as the range where atmosphere leads over accuracy.
For this subject, high stylize is not just defensible but close to necessary. The composition depends on a painterly rendering that no photographic reference supports. Asking a model to hold a low stylize value while simultaneously demanding thick oil-paint brushwork produces something that is neither photograph nor painting — and that is usually the least convincing of the three outcomes.
Going high lets the interpretation carry the medium consistently. The trade-off is explicit and worth stating: at 750 you have given up prompt adherence on everything else. The habit folds, the cross pendant and the hand tattoos will vary between generations. This is a look that needs to be found one generation at a time rather than reproduced on command, and once you find it you should lock the seed.
Where this stops working
The smoke is the weakest element. wisps of smoke curling upward catching neon light asks for a translucent volume with a light source behind it, which is a hard problem in any medium. Expect it to render as ribbon-like strokes rather than as smoke, particularly when it crosses the bright halo where it will usually either vanish or turn into opaque smoke.
Second, text in the background will not be legible. The graffiti is described as including teal Chinese character in lower right corner, and generated characters are texture, not language. If a specific character matters, it has to be added afterwards.
Third, and the most likely cause of disappointment: the skin will not be both painterly and hyper-realistic. Whichever term the model resolves in your generations becomes the look. The useful move is to stop writing both and decide which one you want — thick oil paint texture, visible brushstrokes, canvas grain for a painting, or drop the brushwork words entirely for a render.
What makes this prompt worth reading is not the subject. It is that it shows what happens when a painterly rendering and a photographic one are asked for together, and that the high stylize value is not decoration but the setting that makes the contradiction resolve.
Label: Cinematic Fashion Poster
Key Principle: When a prompt mixes painterly and photographic instructions, the stylize value is what decides which one wins. Painting holds only at high stylize; asking for both at a low value produces neither, and the honest move is to pick one.
Comments
Post a Comment