A freckled young woman with enormous expressive teal eyes and wild honey-blonde hair piled in a messy topknot, wearing a cozy sage-green cable-knit sweater, holds a Cavalier King Charles Spaniel puppy with chestnut-and-white patches and comically oversized amber eyes. Both subjects press cheek-to-cheek in an intimate selfie pose, the woman's left cheek against the puppy's right. The woman forms a surprised "O" mouth shape while the puppy's mouth hangs open in happy panting, pink tongue visible, ears flopped back in excitement—not mirroring but complementary expressions. Golden afternoon light at 45 degrees filters through out-of-focus birch trees behind, creating circular bokeh orbs 20-40% larger than eye highlights, warm amber light on shadow side, cool teal bounce in fill. Ultra-detailed fur rendering with guard hair separation from undercoat, cable knit showing individual stitch definition with cast shadows in recesses, subsurface skin scattering visible at earlobes and thin skin areas. Pixar 3D aesthetic with controlled stylization—eyes 1.4x natural proportion, head 1.15x, maintaining anatomical logic. Cinematic depth of field: f/2.0 equivalent, focus plane on woman's near eye and puppy's near eye with progressive falloff, vertical selfie composition with subjects occupying 75% frame height, 8K render, Arnold render engine quality, ZBrush sculpt detail with tertiary micro-displacement --ar 9:16 --style raw --v 6
Quick Tip: Click the prompt box above to select it, then press Ctrl+C (Cmd+C on Mac) to copy. Paste directly into Midjourney, DALL-E, or Stable Diffusion and customize to make it your own!
The failure mode for this prompt class is predictable and it happens within the first sampling pass. When the stylization ratios are left implicit, the model reads "Pixar 3D" as a permission slip rather than a constraint. The result: the woman's head inflates to roughly 1.6x natural proportion, the eyes push past 2x, and the puppy's muzzle collapses into a single sphere. The image stops reading as a stylized portrait and starts reading as a Funko Pop with skin texture. The cheek-to-cheek contact point, which is the entire emotional anchor of the composition, dissolves into two overlapping blobs because neither subject has a defined jaw plane anymore.
A second failure appears in the lighting logic. Without an explicit angular value, "golden afternoon light" gets interpreted as flat ambient warmth. The model removes the shadow side entirely, and the teal bounce fill disappears. What remains is a uniformly lit face with no subsurface scattering visible at the earlobes, because subsurface scattering requires a terminator line to show through thin skin. No terminator, no glow. The freckles also flatten into a uniform noise layer instead of sitting on top of translucent skin.
This is why the approach here changed. Instead of describing the aesthetic and hoping the ratios land, the prompt now specifies the numeric relationship between stylized and natural proportion. That single change, combined with a hard focus plane and a defined light angle, removes most of the variance. The rest of this breakdown covers the mechanism in order: what the stylization ratio actually controls, how the model resolves the two-eye focus plane, and where the prompt breaks at the edges.
What the Stylization Ratio Actually Controls
Eye proportion and head proportion are not decorative numbers. In a 3D-rendered portrait, they determine the perceived age of the subject and the perceived camera distance. Eyes at 1.4x natural proportion read as youthful and slightly wide-angle, which matches the selfie framing. Head at 1.15x reads as a mild camera-near distortion rather than a stylistic exaggeration. Push either value past roughly 1.5x and the model stops treating the subject as a character and starts treating it as a toy, which changes the fur rendering, the skin shading, and the way the sweater sits on the shoulders.
The ratio also controls the relationship between the two subjects. A 1.15x head on the woman against a natural-proportion puppy head creates a size hierarchy that feels correct for a selfie held at arm's length. If the head ratio drifts upward, the puppy shrinks relative to the frame and the cheek contact point moves off-center. If it drifts below 1.1x, the woman reads as an adult in a photograph and the puppy reads as oversized, which breaks the intimate framing entirely. The window is narrow for a reason.
Anatomically, the ratio interacts with the jaw and cheek plane. At 1.15x, the model still resolves a defined zygomatic arch and a jawline that can press against the puppy's skull. Above that, the jaw softens into a curve and the contact point loses its edge. The same principle applies to stylized portraits with exaggerated facial features, where the difference between intentional and broken comes down to whether the underlying skull structure survives the exaggeration. The technique is documented in more detail in this breakdown of dramatic feathered portraits, where the same ratio discipline governs hair rendering.
For the puppy, the stylization applies only to the eyes. Amber eyes at 1.6x natural proportion on an otherwise anatomically correct spaniel head produce the comic effect without turning the animal into a plush toy. The chestnut-and-white patches stay in their breed-correct positions, the ears flop back with visible cartilage structure, and the open mouth shows a defined tongue and upper palate rather than a pink shape. That contrast, realistic body plus oversized eyes, is what sells the "whimsical" label without losing render credibility.
How the Model Resolves the Two-Eye Focus Plane
The focus instruction in the prompt specifies two points: the woman's near eye and the puppy's near eye. In a real camera, a single plane can only be perfectly sharp at one distance. The model handles this by treating the two eyes as a shallow band, then falling off progressively toward the ears, the sweater, and the background. This is why the f/2.0 equivalent value matters. At f/4.0, the depth of field becomes wide enough that the birch trees behind the subjects start resolving into shapes instead of bokeh orbs, and the composition loses its separation.
At f/2.0, the falloff is steep enough that the sweater's cable knit is sharp at the shoulder but softens at the elbow. That gradient is what signals "photographed" rather than "rendered flat." The model uses the focus plane to decide where to allocate detail density, and the tertiary micro-displacement from the ZBrush reference only appears inside that band. Outside it, the render simplifies. The result is a portrait where the freckles and the individual guard hairs on the puppy's muzzle are resolved, while the background falls into circular bokeh orbs sized 20-40% larger than the eye highlights.
The 45-degree light angle is the second half of this mechanism. At 45 degrees from the subject's front-left, the shadow side of the face receives the cool teal bounce, which sits opposite the warm amber key on the lit side. That temperature split is what makes the subsurface scattering visible at the earlobe and the thin skin beside the nose. Remove the angle and the split disappears. The same light logic applies to cinematic character work, as covered in this stop-motion gothic character prompt, where a defined key direction carries the entire mood.
The bokeh sizing is not arbitrary either. Orbs 20-40% larger than the eye highlights create a background that feels physically closer to the lens than the subjects, which reinforces the selfie framing. Smaller orbs read as distant foliage and flatten the depth. Larger orbs read as foreground obstruction and clutter the frame. The model will produce either extreme if the ratio is left unstated, which is why the prompt locks it.
Edge Cases Where the Prompt Breaks
The first edge case is fur detail at the cheek contact point. When two surfaces meet, the model has to decide which one owns the boundary. With a realistic spaniel head pressed against a stylized human cheek, the fur can either render as individual guard hairs overlapping the skin, or as a single soft mass. The prompt specifies guard hair separation from undercoat, which pushes the model toward the first behavior, but only if the focus plane includes that contact point. If the focus drifts to the woman's far eye, the fur at the contact point softens and the cheek-to-cheek pose loses its tactile quality.
The second edge case is the cable knit. Individual stitch definition with cast shadows in the recesses requires the model to treat the sweater as a geometric surface, not a texture map. If the sage-green value is described too loosely, the model applies a flat color and the stitches disappear. The fix is to anchor the knit to the shoulder and chest planes, which the 75% frame-height composition already does. Subjects occupying 75% of the vertical frame keeps the shoulders inside the sharp band, so the knit resolves where it is visible.
The third edge case is the mouth expressions. A surprised "O" on the woman and an open pant on the puppy are complementary, not mirrored. Models default to mirrored expressions when two faces appear in the same frame, because the training data is full of group selfies where everyone smiles. Specifying the divergence explicitly, "O" shape versus open pant with visible tongue, is what prevents the model from giving both subjects the same mouth. If the prompt omits this, both faces smile and the image loses its comic timing.
Render engine references like Arnold and ZBrush are not literal calls. They function as quality anchors that push the model toward denser micro-detail and more physically plausible shading. The same applies when comparing outputs across Midjourney and Leonardo, where engine-style tokens shift detail density even though neither platform runs the named renderer. Treat them as density instructions, not software.
Before accepting any output from this prompt, run the following checks. Each takes seconds and catches the failure modes described above.
1. Measure the eye-to-head ratio. Eyes should sit near 1.4x natural proportion, head near 1.15x. If the head reads past 1.3x, the render has drifted into caricature.
2. Check the cheek contact point. The fur boundary should show individual guard hairs, not a soft mass. If it is blurred, the focus plane missed the contact point.
3. Verify the light split. One side of the face should carry warm amber, the other cool teal. A flat, uniform face means the 45-degree angle was ignored.
4. Inspect the cable knit at the shoulder. Individual stitches with cast shadows must be visible. A flat green surface means the knit was treated as a texture map.
5. Confirm the expressions diverge. Woman forms an "O," puppy pants with visible tongue. Mirrored smiles indicate the complementarity instruction was dropped.
6. Measure the bokeh orbs against the eye highlights. They should be 20-40% larger. Smaller reads as flat background, larger reads as clutter.
Label: Cinematic
Key Principle: Stylization only reads as intentional when proportions stay within a narrow window: eyes at 1.4x, head at 1.15x. Outside that band, the model defaults to caricature and the anatomical logic collapses.
Comments
Post a Comment