curiousrefuge: Futuristic Spacesuit Cliff Planet AI Art

Consistency is the “holy grail” of AI filmmaking. We tested Nano Banana 2’s ability to generate images using character references for both single and multiple characters, and the results were pretty impressive. Not every character stays perfectly consistent all the time, especially in multi-character scenes, and realism doesn’t hold up in every scenario. But the ability to upload up to 14 reference images opens up real possibilities for narrative storytelling. #ai #generativeai #aifilmmaking #aivideo #aiadvertising

Curious Refuge

@curiousrefuge

INSTAGRAM · 2026-03-13Source

286likes

9comments

Remix This

Recreate with AI Image Generator

Make your own AI viral video

Prompt

[Subject] A polished social-media workflow graphic showing an AI image-generation comparison layout. In the upper panel, display a grid of reference stills featuring the same middle-aged man in a dim domestic interior, wearing a dark blue long-sleeve shirt and appearing in multiple poses and expressions. In the lower panel, show the generated cinematic result: that same man transformed into a serious astronaut standing on a rocky desert cliff in a futuristic spacesuit on an alien world. The layout should also include a prompt card at the bottom describing the transformation from reference identity into the new sci-fi shot.

[Environment] Clean teal-and-dark-green gradient UI background with rounded rectangular cards outlined in a subtle neon-teal stroke. The top card contains the labeled reference-image grid, the middle card contains the generated sci-fi scene, and the bottom card contains a dark prompt box with white text. This should feel like a premium AI tool showcase or product-demo visual, not a raw screenshot.

[Composition/Camera] Vertical infographic composition divided into three stacked sections. Top section: 3x3 reference grid with one close-up labeled “REFERENCE IMAGE.” Middle section: wide cinematic frame labeled “NANO BANANA 2,” showing the astronaut on a cliff with canyon formations and multiple moons in the sky. Bottom section: prompt text block, clearly readable, aligned like a UI instruction panel. Keep consistent rounded corners and margins between all panels.

[Lighting] Soft interface-style ambient glow around the cards, while the generated sci-fi image itself uses cinematic teal-and-sand lighting with atmospheric depth. The reference stills should remain naturalistic and lower contrast, while the output frame should look more epic, stylized, and dramatically graded.

[Style/Rendering] High-end AI product marketing graphic, clean SaaS demo design, cinematic image-generation showcase, editorial interface composition, sharp typography, polished comparison layout, and subtle futuristic UI styling.

[Detail constraints] Preserve the identity continuity between the reference man and the generated astronaut, keep the teal rounded borders intact, maintain clear hierarchy between reference, output, and prompt sections, retain the multi-moon desert canyon environment in the generated image, and avoid clutter or excessive UI chrome that would distract from the transformation concept.

Negative prompt: messy dashboard UI, unreadable text, low-resolution thumbnails, broken identity match, random icons, cluttered buttons, excessive gradients, warped astronaut anatomy, generic space background, overbusy layout, distorted typography, noisy shadows, low-detail canyon scene.

Suggested parameters: stylize 100, quality 1, aspect ratio 4:5, crisp UI layout, readable typography, cinematic output frame, subtle neon edge accents, premium product-demo finish.

Delta prompt strategy: If the layout feels chaotic, add "three clean stacked cards with rounded corners and consistent spacing." If the identity link gets lost, add "same middle-aged male facial structure preserved from reference grid to generated astronaut image." If the sci-fi output feels weak, add "cinematic medium shot of astronaut on a desert cliff with alien moons and canyon depth." If the UI looks generic, add "premium AI tool showcase with teal outlines and dark modern interface styling." If text becomes unreadable, add "large clean sans-serif labels and prompt panel with clear spacing." If the reference section dominates too much, add "reference thumbnails smaller and more subdued than the generated result." If the bottom prompt box disappears, add "dark rounded prompt card with white instructional copy." If the generated image lacks atmosphere, add "teal sky haze, canyon shadows, and cinematic color grading." If the cards lose polish, add "subtle glow edges and refined rounded-corner panel design." If the composition stops reading like a demo, add "before-and-after AI generation product mockup, not a single standalone poster."

This image works because it is not only showing a final picture, it is showing a process. The top grid establishes identity consistency, the middle frame demonstrates the transformed result, and the bottom prompt card explains the instruction that connects them. That three-step structure is what makes the graphic persuasive. It does not ask the viewer to guess what the tool did; it stages the proof visually.

The most useful design choice here is the contrast between the mundane reference context and the cinematic generated output. The source frames are dim, domestic, and observational. The output is expansive, mythic, and heavily graded. Because the same face carries across both sections, the transformation feels impressive rather than arbitrary. This is exactly the kind of structure that sells AI image tooling well: continuity first, spectacle second.

How curiousrefuge Made This Futuristic Spacesuit Cliff Planet AI Art -- and How to Recreate It

Signal	Evidence (from this image)	Mechanism	Replication Action
Process transparency	The graphic explicitly separates reference images, output image, and prompt text	Stepwise layout increases trust because the viewer can inspect the workflow	Design product demos in clear before-and-after stages rather than showing only the result
Identity preservation	The same man appears in both the domestic reference grid and the astronaut output	Continuity makes the transformation feel technically credible	Emphasize consistent facial structure and age across source and generated sections
Spectacle contrast	The output frame shifts from indoor realism to alien-canyon cinematic scale	Large visual gap between source and output makes capability feel more dramatic	Choose reference material that is plain enough to make the transformation legible
Product polish	Rounded cards, teal borders, and clean typography make the layout feel like a tool showcase	Interface refinement makes the workflow feel intentional and premium	Use consistent card styling and restrained UI chrome when presenting model capabilities

Observed Style Choices

Style Choice	Observed Effect
Stacked card layout	Keeps the workflow readable in a fast social-media scroll context
Teal border accents	Creates subtle futuristic branding without overwhelming the images
Muted reference stills	Push the viewer’s attention toward the stronger generated result
Cinematic output frame	Provides the “wow” moment that justifies the workflow graphic
Bottom prompt card	Shows exactly how the transformation was instructed, increasing perceived clarity

Prompt Technique Breakdown

Technique	Why It Matters Here	How To Phrase It
Workflow framing	The image is selling a system, not just a single generated picture	"AI product demo showing reference images, generated result, and prompt panel in one layout"
Identity continuity	The success of the demo depends on believable transformation of the same person	"same middle-aged man preserved across the reference grid and sci-fi output"
Output escalation	The result must feel much more cinematic than the source material	"generated cinematic medium shot in futuristic spacesuit on an alien cliff"
UI restraint	Too much interface chrome would cheapen the presentation	"minimal rounded cards, teal edge glow, clear typography, no extra dashboard clutter"
Caption support	The prompt card explains the transformation and completes the demo logic	"bottom prompt box with concise transformation instruction in readable white text"

Execution Notes

To recreate this kind of graphic, begin by designing the information hierarchy before refining the images. Decide what the viewer should understand in three seconds: reference identity, generated result, and prompt logic. Once that is clear, style the cards consistently and let the output frame carry most of the cinematic excitement.

Most weak versions will either crowd the layout with too much UI or fail to make the identity continuity obvious enough. Fix that by reducing interface noise and ensuring the reference face and output face remain recognizably linked. If the result feels like a random collage, strengthen the stacking logic and make the middle panel visibly superior in cinematic quality. The best version feels like a product demo that earns the viewer’s trust while still delivering spectacle.

How curiousrefuge Made This Futuristic Spacesuit Cliff Planet AI Art -- and How to Recreate It

Observed Style Choices

Prompt Technique Breakdown

Execution Notes

Related AI Generator