Extreme macro profile of a classical female bust sculpted from luminous white porcelain, surface decorated with hand-painted cobalt blue floral motifs in traditional Dutch Delftware style, delicate peonies and trailing vines wrapping organically across curved cheekbones and jawline, liquid glass glaze creating mirror-like specular reflections on high points, museum-quality ceramic photography, soft box lighting from upper left creating subtle rim highlights on nose and lip, shallow depth of field with creamy bokeh isolating form, neutral warm-gray seamless backdrop, hyper-detailed ceramic texture with visible cobalt brushstroke patterns following surface contours, subsurface scattering in thin porcelain areas, 8k resolution --ar 2:3 --style raw --s 750
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!
Approach 1: Material-First Prompting
Material-first prompting front-loads nouns: porcelain, Delftware, ceramic, bust. The model responds by pulling the dominant visual cluster associated with those tokens, which is a smooth white surface with decorative blue patterning. This works, but only up to a point. The output tends to be flatly lit, evenly exposed, and matte, because the model has no geometric instruction telling it where light should strike.
The visible signature is a soft, chalky surface. Highlights are diffuse, roughly 5-10% luminance above midtone. There is no rim light on the nose or upper lip. Cobalt motifs sit flat against the surface rather than following curvature, because nothing in the prompt specifies that the brushwork wraps around jawline and cheekbone. The trade-off is speed and consistency: it will not fail catastrophically, but it will not produce a museum-quality ceramic photograph either.
This is the default for most users. The reason is structural: material tokens are easy to supply, optical tokens require knowing what you want the surface to do. A useful comparison is the way cobalt-decorated porcelain busts are handled in adjacent prompts, where the glaze term is treated as a light modifier rather than a material descriptor.
Approach 2: Optical-First Prompting
Optical-first prompting defines the surface by what light does when it hits it. Instead of "white porcelain," the prompt reads "liquid glass glaze creating mirror-like specular reflections on high points." Instead of "blue flowers," it reads "cobalt brushstroke patterns following surface contours." The model then resolves material from behavior, not from category.
The mechanism is straightforward. Diffusion models weight tokens by their co-occurrence with spatial and lighting cues. "Liquid glass glaze" appears in product photography and ceramic macro datasets almost exclusively alongside directional light, hard specular highlights, and shallow depth of field. The model pulls that entire cluster forward. The result is a surface with two distinct luminance zones: a mid-tone body around 60-70% and specular peaks near 95-100%, which is what makes fired porcelain read as fired porcelain.
The visible signature includes a defined rim highlight along the nose ridge, a compressed highlight on the upper lip, and a bright edge on the brow. Shadows deepen under the jaw. Cobalt motifs gain a subtle bleed at their edges, mimicking underglaze absorption. The trade-off is that without an explicit backdrop and depth cue, optical prompts can produce a cluttered or over-bright frame, since the model pushes light intensity upward. Anchoring with "neutral warm-gray seamless backdrop" and "shallow depth of field with creamy bokeh" restores control.
Approach 3: Hybrid Parameters and the Specular Ceiling
The hybrid approach combines material nouns with optical verbs and explicit numeric parameters. The reference render uses soft box lighting from upper left, which places the primary highlight on the left cheekbone and a secondary rim on the nose and lip. This is not incidental. Upper-left soft box is the default studio key for ceramic and glass product work because it produces a single readable highlight per form and leaves the right side of the face in 15-25% shadow for depth.
Subsurface scattering is the second lever. Thin porcelain areas, especially the ear, nostril edge, and lip line, transmit light rather than reflect it. Adding "subsurface scattering in thin porcelain areas" forces the model to brighten those zones slightly and soften their edges. Without it, they render as opaque. With it, they read as a few millimeters thick, which is the actual wall thickness of slip-cast porcelain.
Brushstroke direction is the third lever. Specifying that cobalt patterns follow surface contours rather than sit as a flat decal is the difference between hand-painted Delftware and printed transfer. The model resolves this through the phrase "hand-painted cobalt blue floral motifs" paired with "wrapping organically across curved cheekbones and jawline." The word "wrapping" carries a spatial constraint the model applies to the pattern generator.
For adjacent workflows where surface behavior dominates the composition, the same optical-first logic appears in organic product photography prompts, where glaze, moisture, and sheen terms do the same work as "liquid glass glaze" does here.
Internally, Midjourney handles this cluster well under --style raw, because raw reduces the model's tendency to stylize surfaces toward illustration. Under default stylization, porcelain drifts toward matte painterly rendering. Under raw with --s 750, the surface keeps its photographic specular behavior. Aspect ratio 2:3 matches the vertical bust composition and prevents the model from padding the frame with unnecessary environment.
Parameter stacking matters here. Each of the three approaches can be run at the same seed, and the difference is measurable in highlight distribution. Material-first produces a histogram with a narrow peak around midtone. Optical-first broadens the histogram toward the highlights. Hybrid clips the specular peaks at 98-100% and keeps the midtone band intact. The cobalt saturation also shifts: material-first yields cobalt around 60% saturation, optical-first pushes toward 85%, and hybrid holds 70-75% with visible brushstroke edge variation.
What differentiates the three approaches is not quality but control. Material-first is a starting point for discovery. Optical-first is for when the surface behavior matters more than the object. Hybrid is for when both need to be pinned down and the render must be reproducible. Choosing between them depends on whether the porcelain is the subject or the support for something else in the frame.
One more variable worth isolating is backdrop luminance. A neutral warm-gray backdrop at roughly 55-65% luminance keeps the porcelain from being read as pure white and preserves the subtle warm cast of fired clay. A pure white backdrop collapses the porcelain into the background and removes the specular contrast that makes the glaze visible. This single change accounts for a large share of the difference between a render that reads as a photograph and one that reads as a mockup.
The boundary appears when the render stops reading as a photograph of a physical object and starts reading as a digital illustration of one. The signal is specular behavior: if highlights spread across surface area rather than concentrating at high points, the model has abandoned optical logic and reverted to material-token defaults. At that point, adding more material descriptors does nothing. The fix is to re-specify the light source, its direction, and the glaze's reflective character, because the surface has to be redefined as an optical event before the material vocabulary will hold.
Label: Product
Key Principle: Treat the porcelain surface as a lighting problem, not a texture prompt. Specify the specular source (soft box, upper-left) and glaze behavior; otherwise the model defaults to matte plastic.
Comments
Post a Comment