Extreme close-up portrait, dark skin with deep melanin-rich undertones, intense ice-blue irises with visible limbal rings framed by red and cream striped knit balaclava mask, weathered hands with dirt accumulation in creases and faded black ink tattoos gripping mask fabric, thorny crimson vine crown woven through short dark hair, single hard light source from upper left at 45-degree angle creating deep chiaroscuro shadows and specular highlights on skin, hyper-realistic skin texture with visible pores, sebum sheen, and fine expression lines, coarse ribbed wool texture with individual fiber detail and slight pilling, shallow depth of field f/2.8 with razor-sharp focus on eyes, pure black negative space background, cinematic moody atmosphere, photorealistic 8K detail, Hasselblad X2D 100C medium format, 90mm lens --ar 9:16 --style raw --s 250 --no smooth plastic skin, soft fill light, blurred knit
Click the prompt box above to select it, then press Ctrl+C (Cmd+C on Mac) to copy. Do not drop the --no clause. It names the three things most likely to go wrong and costs nothing to keep.
This is the most complete prompt in this set, and it is worth reading it as a template rather than as a one-off. It names a real camera, a real light position, three different materials with three different behaviours, a specific aperture, and a --no clause that removes exactly the failure mode most likely to appear.
That combination is what makes it worth using as the reference example for portrait prompts in general.
What the camera reference actually buys you
Hasselblad X2D 100C medium format, 90mm lens is not decoration. Medium format is not just a larger sensor; it changes the look in specific ways that the model has learned.
The square-ish sensor those bodies use produces a different relationship between subject and background than a full frame lens does, and the file characteristics of a large sensor tend toward a rendering with more depth in the shadows and a particular kind of micro-contrast. Naming the body and the lens together gives the model a coherent cluster to match rather than a generic "professional camera" signal.
The editorial prompting guide makes the general point in unusually direct terms: pairing a lighting description with a specific camera and lens reference is the single biggest jump from an AI-looking image to a photographic one. It is not decoration, and this prompt is a good demonstration of the mechanism.
Its advice is also worth taking literally on the lens question. It notes that lens choice governs compression — 35mm for environmental work, 80–110mm for portrait isolation, 24mm for distorted wides. At 90mm this prompt sits squarely in the isolation band, which is right for a face framed this tightly.
One hard light at 45 degrees, and why the angle is specified
single hard light source from upper left at 45-degree angle creating deep chiaroscuro shadows and specular highlights on skin
Three things are packed into that clause and each one is doing measurable work. "Single" means no fill, which is what produces deep shadow rather than grey. "Hard" means a small source, which produces a shadow with a defined edge. "45 degrees" places the terminator between the light and shadow sides of the face at a position that keeps both readable.
That guide lists split lighting as a named setup at 90 degrees with half the face in shadow, and butterfly as a key directly above the lens. This sits between the two, which is a classic portrait position: enough angle to model the cheekbone, not so much that half the face disappears.
It also advises naming the contrast ratio in words, which this prompt does implicitly through deep chiaroscuro shadows. And it lists the generic words to avoid — "good lighting", "professional lighting", "beautiful light" — because a term with no decision attached leaves the model to default to something flat.
The specular highlights in that line are what sell the skin. Oily skin under a hard source produces small bright points, and those points are a strong cue that a face has been photographed rather than modelled.
Three materials, three different behaviours
This is the part that makes the prompt good rather than merely long. Each material is described by what it does.
Skin: visible pores, sebum sheen, and fine expression lines. The sebum sheen is the important term, because it is what interacts with the hard light. Matte skin under a hard source looks like plaster; skin with an oil layer produces the specular points already mentioned. The three terms are consistent with each other rather than contradictory.
Knit: coarse ribbed wool texture with individual fiber detail and slight pilling. Pilling is the detail that does the work — small bobbles on a wool surface are what a model almost never produces unless asked, and their presence is immediately readable as knitwear rather than as a texture fill.
The mask itself: described by construction rather than by name. A ribbed knit balaclava in two colours has a specific silhouette that it either gets or does not.
The photographer's guide argues this principle in general terms, recommending that prompts describe how fabric behaves — silk falls, crepe holds, taffeta stands away from the body — because behaviour carries more information per word than a material name.
The hands are doing narrative work
weathered hands with dirt accumulation in creases and faded black ink tattoos gripping mask fabric
This is the strongest single line in the prompt and it is easy to overlook. The hands are what turn a portrait into a scene: they tell you the mask is being put on or taken off, that the person has been doing something physical, and that dirt sits in creases rather than on the surface.
"Accumulation in creases" is precise physical reasoning. Dirt collects where surfaces meet and rub, not on open areas. Getting that detail right is what separates believable hands from a generic grasping pose.
Where a tattoo crosses a knuckle it will usually break up, because the model does not have a reliable sense of how ink wraps a curved surface. That is worth knowing before you spend attempts on it.
The --no clause is the most useful line in the whole prompt
--no smooth plastic skin, soft fill light, blurred knit
This deserves more attention than it usually gets, because it is the most efficient line in the entire prompt.
Each term targets a specific likely failure. "Smooth plastic skin" is what a diffusion model produces by default, because that is its most common prior for a face. "Soft fill light" negates the hard key explicitly. "Blurred knit" protects against the wool texture dissolving, which is what tends to happen when the aperture is wide and the focus is on the eyes.
Note that all three are descriptions of the thing to avoid rather than instructions to do the opposite. That distinction matters: negative prompts work best when they name the unwanted visual directly, and several community guides treat --no as among the more reliable tools for pushing away model defaults.
The stylize setting and what it costs here
--s 250 again sits in the middle band rather than at either extreme.
Midjourney's stylize documentation gives the range as 0 to 1000 with a default of 100, describing low values as sticking to the prompt's facts and high values as allowing free interpretation. The print-style guide splits that into working bands, with 50 to 150 for structural accuracy and 300 to 750 for atmosphere-led work.
250 suits this prompt because the composition must stay exact — the eye on a focal plane, the hands in the right position — while the skin rendering benefits from some freedom. If the hands keep drifting out of position, drop to 100. If the skin looks lifeless, raise it.
On --style raw: it matters here more than the stylize value, because the default beautification pass is exactly what produces the smooth skin this prompt is explicitly trying to prevent.
Where this stops working
The stated aperture, f/2.8, fights the stated texture. At f/2.8 on a 90mm lens the depth of field is shallow, and shallow focus dissolves the very micro-detail the prompt asks for — individual wool fibres, pores, fine expression lines. In practice the eyes stay sharp and everything else softens, which is usually what you want, but the pilling will frequently be the first thing to go.
Second, ice-blue irises with visible limbal rings on dark skin with deep melanin-rich undertones is an unusual and deliberate combination. It tends to hold, because high contrast between iris and skin is exactly the kind of relationship the model has learned. But the limbal ring specifically may soften or vanish.
Third, a thorny vine crown woven through hair is fine, but vines crossing the mask edge will usually blend with the knit, because both are warm-toned and the model has no reliable separation between them.
What makes this prompt worth copying is not the subject. It is that every material has a behaviour, the light has a position and a size, the camera has a body and a focal length, and the failure modes have been named and excluded.
Label: Assets Cinematic Fashion
Key Principle: Name the failure modes. A --no clause that names the exact unwanted visual — smooth plastic skin, soft fill, blurred knit — is the most efficient line in any portrait prompt, because it removes what the model defaults to.
Comments
Post a Comment