I looove her songs!!! thx @oshuclips for stopping me 🎧⚡️❤️ #oshuclips

imma

@imma.gram · Digital creator

INSTAGRAM · 2024-07-12Source

-1likes

25comments

Remix This

Prompt

HIGH-GRANULARITY INVENTORY

Overall layout
- Format: vertical 9:16 frame with thick black padding/margins around a centered collage
- Collage: two vertical panels side-by-side (left panel slightly wider than right), separated by a thin seam

Left panel (portrait / digital human)
Subject(s)
- Count: 1
- Type: female-presenting digital human / 3D CG avatar, stylized photoreal
- Age appearance: early 20s
- Pose: close-up / head-and-shoulders, facing slightly toward viewer-right, eyes looking forward
- Expression: neutral/calm, lips closed
- Face details: smooth skin, subtle blush, soft pink lips, natural brows, minimal eyeliner, symmetrical features
- Hair: blunt chin-length bob with straight bangs; saturated hot pink/magenta color; smooth CG hair strands with soft highlights
- Accessories: none visible

Clothing & text
- Top: black crew-neck T-shirt
- Printed text on shirt: red arched letters reading “VIRTUAL HUMAN” (curved across chest)

Environment & lighting (left)
- Indoor background, heavily blurred (bokeh-like), warm-beige tones; indistinct poster/art behind
- Lighting: soft even key from front-left; gentle shadows; clean, studio-like face shading

Right panel (street interview / park)
Subject(s)
- Count: 2 foreground people
- Interviewer: male-presenting adult, average build, wearing a bright red baseball cap, white short-sleeve shirt, dark shorts, white socks, sneakers; leaning forward holding a handheld microphone toward the other person
- Interviewee: female-presenting adult with short colorful hair (pink with bluish tint), wearing a white T-shirt with a graphic print, dark skirt/shorts, white socks; mid-step walking; body angled toward viewer-left

Environment & background (right)
- Location: sunny public park
- Ground: green grass lawn
- Background: large leafy trees casting shade; scattered people sitting/standing in the distance
- Lighting: bright daylight, mild shadows; high saturation greens

On-image caption (right)
- Text: “ARE YOU” in all caps
- Placement: bottom center of right panel
- Style: bold condensed sans-serif, bright yellow fill with dark outline/shadow (meme subtitle style)

Composition
- Left panel is a clean close-up; right panel is a medium-wide documentary shot
- Visual contrast: polished CG face vs candid real-world interview moment

Color palette
- Dominant: black + hot pink (left), bright greens + blue sky light (right)
- Accents: red cap, yellow caption text

Image style
- Split-screen social video thumbnail / meme collage; mixed media (CG portrait + live-action photo)


MASTER PROMPT (EN)

[Layout]
Create a vertical 9:16 split-screen collage centered inside thick black margins. Two vertical panels side-by-side: left panel slightly wider than right.

[Left Panel — Subject]
A close-up head-and-shoulders portrait of a female-presenting stylized photoreal digital human (3D CG avatar), early-20s appearance, calm neutral expression, smooth skin, soft pink lips, natural brows. Hair is a blunt chin-length hot-pink bob with straight bangs, smooth glossy highlights.

Wardrobe: black crew-neck T-shirt with red arched text on the chest that reads exactly: “VIRTUAL HUMAN”.

Left background: softly blurred warm indoor environment with beige tones and indistinct poster/art shapes.

Left lighting: soft even studio-like key light from front-left, gentle shadows, clean shading.

[Right Panel — Scene]
A sunny park street-interview moment on a grassy lawn with leafy trees in the background and a few small people sitting/standing in the distance. Foreground includes two people: (1) a male interviewer wearing a bright red baseball cap and a white short-sleeve shirt, leaning forward while holding a handheld microphone, (2) a female interviewee with short colorful hair (pink with a bluish tint) wearing a white graphic T-shirt and a dark skirt/shorts, mid-step walking.

Right lighting: bright daylight with mild shadows, vibrant green grass.

[Right Panel — Text]
Add subtitle-style text at the bottom center of the right panel only, in bold condensed sans-serif, bright yellow with dark outline, reading exactly: “ARE YOU”.

[Detail constraints]
Keep the split-screen structure, black margins, the left shirt text “VIRTUAL HUMAN”, and the right subtitle text “ARE YOU”. Do not add extra captions, watermarks, or logos.

Negative prompt
watermark, logo, extra captions, misspelled text, wrong text, text covering faces, extra panels, wrong aspect ratio, crowded background, indoor interview scene, night lighting, heavy grain, low-res face, uncanny eyes, distorted hands, extra limbs

Suggested parameters (starting points)
- Aspect ratio: 9:16
- Steps: 30–50
- CFG / guidance: 5.5–7.5
- Sampler: DPM++ 2M Karras (or equivalent)
- Style strength: medium
- Seed: lock once split-screen geometry is correct (e.g., 514112903)

Delta prompt strategy (top drift risks + corrective micro-prompts)
1) Split-screen geometry drifts → “two vertical panels side-by-side, left panel slightly wider, centered collage with thick black margins”
2) Left shirt text wrong → “red arched text on shirt reads exactly: VIRTUAL HUMAN”
3) Right subtitle text wrong → “bottom-center yellow subtitle on right panel reads exactly: ARE YOU”
4) Left becomes too realistic or too cartoon → “stylized photoreal 3D digital human, smooth skin shading, crisp eyes”
5) Right loses park setting → “sunny green park lawn, leafy trees, people sitting in distance”
6) Red cap missing → “interviewer wearing bright red baseball cap holding a microphone”
7) Lighting mismatch → “left soft studio lighting; right bright outdoor daylight”
8) Black margins disappear → “collage inside thick black padding around the image”
9) Panels swap or blend → “hard vertical seam separating panels, no blending”
10) Background too busy → “right background simple park, few distant people, mild blur”

How imma.gram Created This Virtual Human Interview AI Art

This image feels like a thumbnail for a story you want to finish. One side is polished and controlled (a virtual human portrait). The other side is chaotic and social (a park interview with a mic). That contrast is the hook.

Why this goes viral: the format creates a question before the words do

People stop scrolling when they sense two realities colliding. The left panel is a clean, studio-like digital portrait—smooth skin, perfect hair, and a shirt that literally labels the concept: “VIRTUAL HUMAN.” The right panel is the opposite: a handheld-feeling street interview in a sunny park, with a microphone pointed mid-conversation. Your brain reads it as a before/after, an explanation, or a reveal.

The yellow subtitle fragment “ARE YOU” is doing a very specific job. It’s not a complete sentence, which makes it feel like you paused a video at the most important moment. That incomplete line creates an open loop: Are you what? Are you real? Are you the virtual human? Are you a fan? The viewer has to imagine the next word—and that tiny act increases dwell time.

Finally, it’s a “proof + premise” thumbnail. The premise is the identity claim (virtual human). The proof is the social context (someone stopping you with a mic in public). Together, they imply credibility and a story worth watching, even if the post is just one frame.

Signal Table

Signal	Evidence (from this image)	Mechanism	Replication Action
Mixed-media contrast	Polished CG portrait vs candid park interview	Collision of worlds creates curiosity and “what’s happening?” energy	Pair a clean studio panel with a messy real-world panel; don’t make both polished
Open-loop subtitle	Single fragment: “ARE YOU”	Paused-at-the-best-moment effect increases watch intent	Use 1–3 word subtitle fragments (“ARE YOU”, “SO YOU”, “WAIT…”) on frame 1
Identity labeling	Shirt text: “VIRTUAL HUMAN”	Instant premise comprehension at thumbnail size	Put the concept label on wardrobe or a prop; keep it readable and simple
Social proof cue	Microphone + public setting	Signals relevance: “this is worth stopping for”	Add a mic/interviewer silhouette to imply public interest

Use cases & transfers

Best-fit scenarios

Creator identity reveals: Use split-screen to show “who I am” vs “how people react.” Keep the left panel controlled and branded.
Launch teasers: Left is product/character; right is reaction clip. The format implies proof without needing long text.
Street interview series: Keep the park panel consistent and rotate the left panel topic (music, AI, fashion, travel).
Collab framing: Left is the collaborator; right is the moment you met or got asked about them.

Not ideal

Single-message announcements: Split-screen implies story; if there’s no payoff, viewers feel baited.
Highly sensitive topics: Street-interview energy can trivialize serious messages.
Dense education: This is a hook format; put the lesson in the follow-up frames.

Transfers (3 recipes)

Transfer 1: “Studio vs backstage” artist reveal
- Keep: split-screen, concept label on wardrobe, incomplete subtitle fragment
- Change: right panel park → backstage hallway with crew blur
- Slot template: “Left: clean portrait with {concept_label}; Right: candid {behind_the_scenes_scene} with mic or crew; subtitle fragment: {2_words}”
Transfer 2: “AI tool vs real reaction”
- Keep: proof + premise structure
- Change: left panel becomes a product UI screenshot (simplified); right panel remains street reaction
- Slot template: “Left: {tool_result} labeled; Right: {reaction_scene}; subtitle fragment: ‘ARE YOU…’ style open loop”
Transfer 3: “Fan question” series thumbnail
- Keep: yellow subtitle style on the right, clean branding on the left
- Change: subtitle fragment rotates (“HOW DID”, “WHY DO”, “CAN YOU”)
- Slot template: “Split-screen: branded portrait + public question moment; right subtitle: {question_fragment}”

Aesthetic read: two lighting worlds, one attention funnel

The left panel is designed to be frictionless: soft even light, smooth skin, controlled color, clean crop. It reads as “official.” The right panel is designed to feel alive: daylight greens, motion energy, and a microphone that implies conversation. The black padding around the collage makes it feel like a captured moment from a vertical video edit—native to the platform.

Color is a quiet hero here. Pink hair links the panels, while the red cap and yellow subtitle give the right panel a pop that feels like entertainment content. The result is a thumbnail that’s both branded and dynamic.

Observed → Recreate (evidence table)

Observed	How to recreate it (prompt + knob)
Split-screen inside black margins	Specify “collage centered with thick black padding”; lock panel widths and seam
Concept label on wardrobe	Add short readable shirt text (“VIRTUAL HUMAN”) and keep it unambiguous
Street interview proof	Include “handheld microphone” + public park background for instant context
Incomplete subtitle fragment	Use 1–3 words in yellow outlined subtitle font; place bottom-center of right panel
Panel link via hair color	Repeat one signature feature across panels (hair color, outfit element, icon)

Prompt technique breakdown (control manual)

Prompt chunk	What it controls	Swap ideas (EN, 2–3 options)
Layout: split-screen + margins	Thumbnail readability and “native edit” feel	“two-panel split-screen” / “three-panel collage” / “picture-in-picture”
Left panel: branded portrait	Authority and identity clarity	“studio portrait” / “clean gradient background” / “close-up beauty shot”
Right panel: public proof scene	Social energy and relevance	“park interview” / “street at night” / “event crowd blur”
Wardrobe label	Instant premise comprehension	“VIRTUAL HUMAN” / “AI GIRL” / “DIGITAL ID”
Subtitle fragment	Open loop and watch intent	“ARE YOU” / “SO YOU” / “WAIT…”
Linking motif	Whether the two panels feel related	“same hair color” / “same accessory” / “same logo color”

Quick prompt skeleton

Vertical 9:16 split-screen collage inside thick black margins. Left: clean studio portrait of {virtual_subject} wearing a shirt that says {label_text}. Right: candid {public_scene} with an interviewer holding a microphone. Add yellow subtitle fragment at bottom right panel: {2_words}.

Remix steps: make a whole series from one split-screen template

Baseline lock (lock these first)

Geometry: consistent panel widths + black margins + seam
Premise clarity: one readable label on wardrobe or prop
Proof scene: microphone + public space + daylight (simple, recognizable)

One-change rule

Change only one knob per upload: either the wardrobe label (premise), the subtitle fragment (hook), or the right-panel scene (proof). Keep the rest identical so the audience learns your format.

Example 4-step iteration sequence

Run 1: Nail the split-screen geometry and margin thickness.
Run 2: Lock the left portrait lighting and the shirt label readability.
Run 3: Add the right interview scene and keep it slightly messy (real-feeling).
Run 4: Test 5 subtitle fragments (“ARE YOU”, “WAIT…”, “SO YOU”) and keep the best performer.