Extreme close-up portrait, woman with pale skin, round metallic-framed sunglasses with iridescent yellow-green gradient lenses showing digital grid reflections and a visible green eye with dramatic black eyelashes behind the left lens, glossy dark navy blue lips slightly parted showing teeth, thick dark eyebrows, short black choppy bangs, black choker with white text "GRUAT", cool blue cinematic lighting from left creating soft shadows on the right side of face, shallow depth of field, hyper-realistic skin texture with visible pores and subtle freckles, cyberpunk aesthetic, fashion photography style, moody atmosphere, high contrast --ar 9:16 --style raw --s 250
Click the prompt box above to select it, then press Ctrl+C (Cmd+C on Mac) to copy. The --s 250 value is deliberate here: if the eye behind the lens keeps disappearing, lower it to 100.
This prompt has a fixed point of view: the left lens of the sunglasses is transparent enough to see a green eye behind it. Everything else — the iridescent yellow-green gradient, the digital grid reflections, the navy lips, the short choppy bangs — serves that single image.
That focus is what makes it work, and it is also what makes it fragile. Reflections on a lens and a visible eye behind that lens are two separate physical things, and the model has to satisfy both at once.
Pick the lens before you pick the words
An extreme close-up is a lens decision. The prompt does not name a focal length, which is the main thing missing from it, and that omission matters more here than in any other prompt in this set.
A guide to editorial-quality image prompts is explicit that lens choice determines compression: 35mm for environmental shots, 80–110mm for portrait isolation, 24mm for distorted wides. For a frame this tight, a long lens is what you want, and it should be named. 105mm macro at f/2 would give you the shallow depth of field and the working distance this composition needs.
The same guide's most useful warning for close-ups is about stacking contradictions. It advises against piling incompatible setups into one prompt, because the model averages them into mush. This prompt asks for shallow depth of field and hyper-realistic skin texture with visible pores and subtle freckles at the same time. Those pull against each other: a very shallow focus dissolves exactly the micro-detail you asked to see. If the freckles keep smearing, that is the conflict, and the fix is a slightly deeper aperture rather than repeating the texture words.
What is actually happening on the lens
The colour on the lenses is not decoration. It is a described optical effect, and describing it as a gradient rather than as a colour is what makes it behave like glass rather than paint.
iridescent yellow-green gradient lenses showing digital grid reflections works because a gradient implies a shift across the surface rather than a flat tint. The eye then reads it as an angle-dependent coating, which is what a treated lens actually looks like.
Two details are doing more work than they appear to. The visible green eye with dramatic black eyelashes behind the left lens is spatially precise — left, behind — and that precision is what gives the model a chance to place it correctly rather than averaging it across both lenses. And glossy dark navy blue lips slightly parted showing teeth adds a specular highlight that the eye needs in order to read the lips as wet rather than as flat colour.
If the eye keeps disappearing, the most likely cause is that the reflection wording is winning over the eye wording. Reduce digital grid reflections to a single pass across the top of the lens and the eye usually comes back.
One light source, and say which side
cool blue cinematic lighting from left creating soft shadows on the right side of face is the lighting decision in this prompt, and it is a real one: a single source at a defined side produces a shadow side with a defined edge.
The same guide catalogues the setups it recommends naming explicitly — split lighting at 90 degrees with half the face in shadow is precisely what this describes — and argues for specifying angle, height, size and hardness each time, plus the contrast ratio in words.
One thing to be careful about: that guide lists "cinematic" among the generic terms to avoid, because a word with no decision attached leaves the model to default to something flat. The word is doing real work in this prompt only because it is immediately followed by a direction and a shadow description. Strip those and the word becomes noise.
Cool light against warm skin is a temperature contrast rather than a lighting choice, and it is what keeps the skin from going grey. If the complexion turns sallow, naming a warm fill from the right side fixes it without touching the key.
The parameter here is deliberate, and it is worth understanding
This prompt sets --s 250, which is a departure from the other prompts in this set. It is not an arbitrary number.
Midjourney's stylize documentation puts the range at 0 to 1000 with a default of 100, and describes low values as sticking to the facts of the prompt while high values allow the model to interpret more freely and stray from the exact details. A woodcut and linocut guide splits that range into working bands, treating 50 to 150 as strict prompt adherence for work where structure must stay accurate, and 300 to 750 as the range for looser, more atmospheric work where mood matters more than precision.
250 sits in the transition between those bands. For this portrait it is defensible: the composition has to stay precise, but the cyberpunk colour treatment is a matter of atmosphere rather than accuracy. If your results are drifting — the eye not where you asked, the grid landing on both lenses — drop to --s 100. If the result is technically correct but lifeless, raise it.
--style raw is the more important of the two here. It suppresses the default beautification pass, which is what removes the waxy skin and the over-smoothed lips that this close framing exposes most easily.
Aspect ratio is a framing decision
--ar 9:16 is a tall vertical crop. For an extreme close-up that reads as a phone-first format, and it suits a portrait crop from the forehead to the chin with room above the head.
It does constrain the horizontal field, though. At 9:16 the frame is narrow, so anything at the sides — the far lens, the falloff of the hair — gets squeezed. If both lenses need to read clearly, --ar 4:5 is the more forgiving choice.
Where this stops working
The honest limit here is the eye behind the lens. Reflections and a visible eye are in tension physically: a curved reflective lens shows whoever or whatever is in front of it, and the model resolves this by simplifying. It usually succeeds on one lens and fails on the other, which is why the prompt specifies the left lens explicitly instead of saying "visible behind the lenses".
Also worth knowing: the text rendered on the choker is generated, not typed. "GRUAT" is a string the model produced as plausible lettering, not a design decision. If it matters, set it in a real design tool afterwards.
The useful framing is that this prompt is not asking for a cyberpunk portrait. It is describing one lens as transparent, one as reflective, and lighting the face from a direction that keeps both readable at once.
Label: Assets Cinematic Fashion
Key Principle: A close-up needs a focal length. Lens choice governs compression and working distance, and this prompt has one structural element to protect: the eye visible behind one lens. Say which lens, and reduce the reflection wording if it disappears.
Comments
Post a Comment