Complete Prompt + Function Map
Here's the full prompt — it looks like just two sentences, but every phrase is an independent "control switch":
A giant [CONTAINER TYPE] with a miniature 3D diorama
inside it, depicting [SCENE DESCRIPTION], studio-lit
with soft shadows, placed on a neutral matte surface.
Highlighting texture, light refraction, and emotional
tone. Hyper-realistic details, macro photography style.
The two brackets [CONTAINER TYPE] and [SCENE DESCRIPTION] are variable slots — fill in container type and internal scene description respectively. Every other phrase controls a specific rendering dimension.
Function map overview:
| Module | Prompt Fragment | Controls | Effect When Replaced |
|---|---|---|---|
| ① Scale anchor | A giant |
Container size relative to frame | tiny → becomes jewelry-scale |
| ② Container | [CONTAINER TYPE] |
Container material and form | droplet / bulb / capsule / bottle |
| ③ Miniature trigger | miniature 3D diorama inside it |
Internal rendering mode | Remove 3D → may become flat decal |
| ④ Scene narrative | depicting [SCENE DESCRIPTION] |
Internal scene content | Fill with any scene |
| ⑤ Lighting | studio-lit with soft shadows |
Light direction and shadow quality | dramatic → becomes theatrical |
| ⑥ Surface | placed on a neutral matte surface |
Physical environment | floating → levitation effect |
| ⑦ Render priority | Highlighting texture, light refraction |
AI's rendering focus | Remove refraction → container turns opaque |
| ⑧ Output standard | Hyper-realistic, macro photography |
Final image style | watercolor → becomes painted |
Word-by-Word Breakdown — Why Each Module Exists
Module ①: A giant — Scale Anchor
giant doesn't just mean "big" — it sets the container's relative scale to the viewer.
AI's scale rendering logic: giant makes the container dominate the frame (60-80%), implying the viewer is observing a large object up close. This scale relationship gives the internal miniature scene more rendering space — AI "packs in" more detail.
| Scale Word | Frame Proportion | Internal Detail | Best For |
|---|---|---|---|
giant |
60-80% | Maximum detail | Showcasing rich internal scenes |
large |
40-60% | High | Balancing container exterior and interior |
small |
15-30% | Medium | Emphasizing container's environmental presence |
tiny |
5-10% | Minimal | Presenting container as jewelry/ornament |
Common mistake: Writing a container without a scale word — AI defaults to medium size, internal scene may lack precision.
Module ②: [CONTAINER TYPE] — Container Material Selector
This first variable slot determines the container's form, material, and optical properties.
Container type directly affects three rendering dimensions: refraction mode (convex lens vs flat glass vs translucent scatter), surface reflection (mirror vs diffuse vs metallic), and internal visibility (fully transparent vs semi-transparent vs tinted).
6 classic containers and their rendering characteristics:
| Container | Refraction Type | Difficulty | Best Scene Pairing |
|---|---|---|---|
glossy water droplet |
Convex lens, slight internal distortion | ★★★★ | Natural/organic scenes |
transparent glass bottle |
Flat, clear internal view | ★★★ | Safe choice for any scene |
vintage light bulb |
Curved + filament interference | ★★★★★ | Industrial/retro scenes |
sleek tech capsule |
Semi-transparent scatter | ★★★ | Sci-fi/futuristic scenes |
crystal sphere |
Spherical omnidirectional refraction | ★★★★ | Fantasy/magical scenes |
cracked hourglass |
Irregular refraction + cracks | ★★★★★ | Time/decay themes |
Interested in glass refraction control inside transparent containers? Our transparent display box 3D guide compares crystal-clear / frosted / tinted glass refraction differences — the same optical physics applied across different container forms.
Module ③: miniature 3D diorama inside it — Miniature Render Trigger
These 5 words are the key to triggering AI's "independent 3D scene inside container" rendering mode.
Each word's rendering function: miniature (scale reduction) + 3D (volumetric rendering, not flat texture) + diorama (scene with foreground-background depth) + inside it (locks spatial relationship inside the container). Remove 3D and the interior becomes flat; remove diorama and it becomes a blurry color blob; remove inside it and the scene may appear outside the container.
Module ④: depicting [SCENE DESCRIPTION] — Scene Narrative Slot
The second variable slot — fill with specific visual descriptions for the internal scene.
Scene description detail level directly determines internal rendering richness:
- Vague:
a forest→ AI renders a few tree outlines, lacks narrative - Medium:
a lush green forest with a river→ basic scene elements present - Precise:
a lush green forest with a winding river, a tiny wooden cabin on the riverbank, morning mist rising between the trees→ AI renders a narrative miniature world
Scene formula: [environment] + [focal element] + [atmosphere] + [light]
3 scene examples:
a snowy mountain village with tiny glowing windows, pine trees heavy with snow, a frozen lake reflecting the stars— winter fairy talea coral reef ecosystem with colorful fish, swaying seaweed, and bioluminescent jellyfish— underwater worlda decaying Gothic cathedral overgrown with vines, shafts of light through broken stained glass— ruin aesthetic
Module ⑤: studio-lit with soft shadows — Lighting System
studio-lit isn't simply "has lighting" — it triggers commercial photography-grade multi-source rendering. It activates a multi-light setup (key + fill + rim light); soft shadows makes shadow edges diffuse softly, implying softbox usage.
Lighting replacement experiments:
| Lighting | Source Type | Shadow Quality | Mood |
|---|---|---|---|
studio-lit with soft shadows |
Multi-source soft | Soft gradient | Professional, calm, commercial |
dramatic single spotlight |
Single spot | Hard shadow | Theatrical, sacred, focused |
natural window light |
Side natural | Directional long | Everyday, warm, lived-in |
backlit with rim glow |
Backlight | Edge glow | Mysterious, ethereal, silhouette |
Module ⑥: placed on a neutral matte surface — Surface Environment
This defines the physical environment where the container exists — not "background" but where the container "stands."
neutral (neutral color, doesn't compete for attention) + matte (matte texture, creates soft contact shadows rather than mirror reflections) + surface (horizontal platform, gives physical support). Replace with floating in a dark void (isolated levitation), resting on wet black marble (mirror reflection), half-buried in white sand (discovered relic feeling), or held by a human hand (adds human scale reference).
Module ⑦: Highlighting texture, light refraction — Render Priority Declaration
This sentence isn't description — it's a rendering weight instruction for AI.
Highlighting tells AI: "Among everything that needs rendering, prioritize allocating computational resources to these items." texture (container surface material detail) + light refraction (light bending through container walls — the core of "container feel") + emotional tone (gives the image emotional direction). If you remove light refraction, the container may become an opaque solid — this is the most easily overlooked yet most essential word in the entire prompt.
Module ⑧: Hyper-realistic details, macro photography style — Output Standard
Hyper-realistic details demands physically correct reflection, refraction, and scattering on every surface. macro photography style triggers not just "close-up" but also extreme shallow depth of field (container edges begin to blur), macro-level detail density (tiny air bubbles in glass visible), and slightly elevated color saturation. Replace with cinematic wide shot for a completely different result — container shrinks, environment expands, shifting from "examining a precise object" to "discovering an object in a scene."
Assembly Order Logic — Why This Sequence Works Best
The word order follows AI's prompt processing priority logic:
- Subject before attributes:
A giant [CONTAINER](subject) →miniature 3D diorama(contents) →depicting [SCENE](details) - Light before surface:
studio-lit(lighting) →placed on surface(environment) - Declaration before standard:
Highlighting ...(render priority) →Hyper-realistic ...(output standard)
Placing Hyper-realistic details first would cause AI to over-focus on global realism while neglecting internal scene details — prompt beginnings typically carry the highest weight. Best practice: subject description first, quality standards last.
3 Variation Experiments
Experiment 1: Same Container, Three Scenes — "Container as Frame"
Fixed container: transparent glass bottle, only swapping scenes:
Variant A: a cherry blossom garden in full bloom, petals drifting, a stone lantern beside a koi pond
Variant B: a volcanic eruption with flowing lava, ash clouds, a tiny village fleeing at the base
Variant C: a cozy bookshop with floor-to-ceiling shelves, a reading cat on a velvet chair, warm lamplight
Finding: The same bottle's color temperature shifts with internal scenes — cherry blossoms tint the glass pink, volcano turns it orange-red, bookshop gives it warm amber. AI "bleeds" internal light source color temperature through the container walls — correct physical refraction behavior.
Experiment 2: Same Scene, Three Containers — "Container as Filter"
Fixed scene: a lush tropical island with palm trees and turquoise water, only swapping containers:
Variant A: A giant glossy water droplet
Variant B: A giant cracked glass sphere
Variant C: A giant amber-tinted vintage perfume bottle
Finding: The water droplet's convex lens magnifies the center and compresses edges; cracked glass projects irregular light-division lines across the internal scene; amber perfume bottle overlays the entire tropical scene with a vintage warm filter. The container isn't just "a shell holding things" — it's an optical filter that transforms the internal world's visual expression.
Experiment 3: Changing Lighting — "Light as the Third Storyteller"
Fixed container and scene (glass bottle + snowy village), only swapping lighting:
Variant A: studio-lit with soft shadows
Variant B: single dramatic spotlight from above
Variant C: backlit with golden rim glow, dark front
Finding: Soft light creates "exhibition mode" — like a museum display, evenly bright. Spotlight creates "sacred mode" — a light beam from above, the bottle as a chosen artifact. Backlight creates "mystery mode" — golden rim glow with dark front, internal scene half-visible. Lighting doesn't change the scene content, but completely transforms its narrative tone.
Test the baseline and all 9 variants across the 3 experiments in nanobanana pro to compare how container × scene × lighting cross-effects reshape the same world.
Common Failures and Fixes
Failure 1: Container Is Completely Opaque — Can't See Inside
Cause: light refraction lacks weight; AI renders the container as a solid object.
Fix: Strengthen transparency — append the container is crystal-clear transparent, the internal scene is fully visible through the glass walls, light refracts beautifully at the container edges.
Failure 2: Internal Scene Is a Flat Decal — No 3D Depth
Cause: 3D diorama trigger isn't strong enough; AI treats the internal scene as "a picture pasted on the container's inner wall."
Fix: Append the miniature scene inside has full 3D depth — foreground elements are closer to the viewer, background elements recede, with visible shadows between layers.
Failure 3: Container and Internal Scene Lighting Don't Match
Cause: External studio lighting conflicts with internal scene sunlight — two contradictory light systems.
Fix: Append the internal scene's lighting is consistent with the external studio light — the main light enters from the same direction. Alternatively, deliberately use contradictory lighting (dark exterior, self-luminous interior) to create "the container holds an independent world" surrealism.
Failure 4: Container Surface Too "Clean" — Looks Like 3D Modeling
Cause: Missing surface micro-imperfections. Real glass/water has bubbles, micro-scratches, light spots.
Fix: Append the container surface has subtle imperfections — tiny air bubbles, micro-scratches catching the light, small condensation droplets on the outside.
Interested in the "micro-imperfection" technique for making AI images look like real photographs? The material realism section of our levitating sliced fruit photography guide discusses the same principle — how small imperfections make AI images look physically real.
Failure 5: Scene Too Complex — Details Collapse Into Color Blobs
Cause: Scene description has too many elements, exceeding the limited rendering capacity inside a container.
Fix: Simplify — a good container scene typically needs only 1 focal element + 1 environmental background + 1 atmosphere element. Example: a lonely lighthouse on a cliff, rough sea below, stormy sky — three elements tell a complete story.
Failure 6: No Refraction Distortion at Container Edges
Cause: AI didn't render light bending through curved glass walls.
Fix: Append the curved glass walls cause visible refraction distortion — the internal scene appears slightly magnified and warped near the edges, with chromatic aberration at extreme angles. Chromatic aberration especially enhances the "viewing through real glass" physical authenticity.
FAQ
Can the container hold "impossible" scenes — like outer space or the ocean floor?
Absolutely — this is the style's most powerful surreal application. depicting a deep space nebula with swirling galaxies inside a water droplet creates "the universe in your palm." AI doesn't check whether the scene could physically exist inside the container — it faithfully renders what you describe. "Impossible" combinations actually produce the highest-engagement content for this style.
Can the container be non-transparent — like ceramic or metal?
Yes, but you need to change how the interior is revealed. Non-transparent containers need openings: a giant ceramic bowl with the top open, revealing a miniature garden inside or a cracked metal sphere with light leaking through the cracks, hinting at a glowing city inside. Refraction is no longer the focus — replaced by "opening composition" or "crack-peek" narrative techniques.
How do I create a series of multiple containers?
Use a row of 3 identical glass bottles, each containing a different miniature world — the left one depicts [Scene A], the middle depicts [Scene B], the right depicts [Scene C]. The key is appending identical container design, identical lighting, identical surface to lock container and environment consistency, letting only internal scenes vary. This "same container, different worlds" series format drives extremely high social media engagement — viewers naturally compare the three worlds.
Can internal scenes include "living" creatures?
Yes — AI can render miniature people, animals, even fantasy creatures inside containers. The key is adding scale anchoring: tiny human figures the size of ants walking through the miniature city explicitly tells AI the relative size between figures and scene. Without scale anchoring, AI may render figures too large (filling the container) or too small (invisible).