本日配信リリース!!! 「EAT THE PAST (English Ver.)」 Lyrics/Music/Arrangement: Iori Kanzaki Localization: Electic Squid / Lachlan Johnson / Hiroki Ueda / kahoca(Empty old City) 先月5月にシカゴで開催された「Anime Central 2025」で初披露した、花譜の代表曲「過去を喰らう」英語版を配信リリース！ぜひお聴きください。

花譜

@virtual_kaf · ai-influencer

INSTAGRAM · 2025-06-10Source

music

3.8Klikes

28comments

Remix This

Prompt

HIGH-GRANULARITY INVENTORY (what’s visible)

Subject(s)
- Count: 1 main subject centered in frame.
- Apparent age range: late-teen to young adult (stylized/anime).
- Gender expression: feminine-presenting.
- Pose: standing, shoulders squared to camera, head facing forward.
- Expression: calm / slightly vulnerable; lips slightly parted.
- Hair: pastel pink hair with straight blunt bangs; two long braids/pigtails draped forward over the chest; subtle cyan rim glow along hair edges.
- Face: simplified anime facial features; small nose; smooth skin; no freckles.
- Eyes: large anime eyes with a vivid red-to-orange iris and a warm yellow inner glow; dark pupil; glossy highlight.
- Accessories: none visible (no earrings/necklace/glasses).

Clothing & materials
- Outerwear: oversized hoodie/jacket in deep navy/teal with red accents/stripes along the shoulders/sleeves; soft fabric with minimal folds.
- Inner layer: light/white top visible at the neckline.

Props/objects
- Background city elements: distant buildings with windows; street-level structures.
- Foreground/near background: a railing/guardrail with vertical bars (left/lower area).
- Street elements: a few cars visible at lower right; street lane/sidewalk hints.
- Practical light: a bright street lamp / light fixture glowing at the top-right corner.
- Text overlay: white serif text that reads “Searching for your smile,” with:
- “Searching” set vertically along the left edge,
- “for your” near top-left,
- “smile,” large and centered-left.

Environment
- Location type: urban street / city block at night.
- Time-of-day cues: nighttime; artificial lighting; cool blue ambience.
- Depth layers: subject in sharp focus; background heavily defocused with bokeh and haze.
- Weather: not clearly raining; slight atmospheric bloom suggests humid/hazy air.

Composition
- Framing: medium close-up portrait; subject fills ~65–75% of frame height.
- Orientation: vertical (phone/reel cover feel), ~9:16.
- Camera height: eye level.
- Angle: straight-on.
- Subject placement: centered; head near upper-middle; negative space on top-left used for text overlay.

Lighting
- Key/fill: soft frontal fill (very low contrast), likely from screen/softbox-like source.
- Rim: cool cyan/blue rim light outlining hair and shoulders (especially on camera-right side).
- Practical: strong white-blue streetlamp glow top-right creating bloom.
- Shadows: minimal; smooth gradients consistent with anime/3D render.
- Color temperature: overall cool (blue) with warm accents in eyes and red wardrobe accents.

Color palette
- Dominant: cobalt/ultramarine blues, cyan glow.
- Accents: pastel pink hair, red wardrobe accents, warm orange/yellow eyes.
- Contrast: medium; lifted blacks; soft bloom.

Image style
- Style: anime-inspired 3D render / cel-shaded look; smooth surfaces; subtle bloom; shallow depth of field.
- Lens characteristics: portrait telephoto feel (compressed background), strong bokeh.
- Sharpness: subject moderately sharp; background very soft/defocused.

──────────────────────────────────────────────────────────────────────────────
MASTER PROMPT (English, recreate ALL details)

[Subject]
Single anime-inspired 3D girl centered in frame, late-teen/young-adult stylized character, facing camera at eye level, calm slightly vulnerable expression with lips slightly parted, large glossy anime eyes with vivid red-to-orange irises and a warm yellow inner glow, smooth pale skin, pastel pink hair with blunt straight bangs and two long braids draped forward over her chest, no jewelry, no glasses.

[Environment]
Nighttime urban street scene, modern city buildings with windows in the distance, street-level structures and a guardrail/railing with vertical bars near the lower-left, a few cars visible at the lower-right, atmospheric haze and bloom.

[Composition/Camera]
Vertical 9:16 portrait, medium close-up, subject fills ~70% frame height, straight-on camera, centered subject, background heavily defocused with strong bokeh, telephoto portrait look (around 85mm), shallow depth of field.

[Lighting]
Cool blue ambient night lighting, soft low-contrast frontal fill on face, distinct cyan/blue rim light outlining hair and shoulders (stronger on camera-right edge), bright practical streetlamp glow at the top-right with soft bloom, minimal hard shadows.

[Style/Rendering]
Anime-style 3D render / cel-shaded aesthetic, smooth gradients, subtle filmic bloom, clean but slightly dreamy look, gentle noise, high-quality render, realistic bokeh.

[Text overlay]
Add white serif typography overlay that reads “Searching for your smile,” with “Searching” arranged vertically along the left edge, “for your” near the top-left, and “smile,” larger and centered-left; keep the text crisp and legible without covering the face.

[Detail constraints]
Do not add or remove objects; keep exactly one subject; keep two braids; keep the streetlamp in the top-right; keep cars faint at lower-right; keep the cool-blue night city bokeh background; keep the same text and layout; keep the subject centered and background blurred.

Negative prompt
extra people, extra faces, duplicate character, missing braids, different hair color, wrong eye colors, open smile, teeth showing, glasses, earrings, hats, busy patterned background, daylight, rain streaks, sharp background, low-res, watermark, logo, misspelled text, messy typography, illegible letters, distorted hands, deformed anatomy, photorealistic human skin, harsh shadows, overexposure, chromatic glitches.

Suggested parameters (starting point)
- Aspect ratio: 9:16
- Resolution: 832x1472 (or 768x1365)
- Sampler: DPM++ 2M Karras (or similar)
- Steps: 32–40
- CFG / guidance: 6–8
- Style strength (if applicable): 0.6–0.8
- High-res fix / upscaler: 1.5–2.0x with mild denoise (0.25–0.35)
- Seed: pick a fixed seed once you like the base (e.g., 284613927)

Delta prompt strategy (10 common drifts + corrective micro-prompts)
1) Text overlay layout drifts → “white serif text: ‘Searching’ vertical left edge; ‘for your’ top-left; ‘smile,’ large center-left; keep text off the face”
2) Hair loses two braids → “two long front braids, both visible on chest, pastel pink, blunt bangs”
3) Eye color becomes generic → “irises red-to-orange gradient with warm yellow inner glow, glossy highlights”
4) Rim light disappears → “strong cyan rim light outlining hair and shoulders, cool neon edge glow”
5) Background becomes sharp → “background heavily defocused, strong bokeh, shallow depth of field, telephoto portrait”
6) Night city turns into generic studio → “night urban street, distant buildings and windows, street-level structures, faint cars lower-right”
7) Streetlamp missing or moves → “bright streetlamp glow at top-right corner with soft bloom”
8) Hoodie colors change → “oversized navy/teal hoodie with red accents/stripes, simple fabric”
9) Expression changes (smiling/angry) → “neutral soft expression, slightly parted lips, calm gaze”
10) Framing shifts off-center → “centered composition, medium close-up, subject fills ~70% height, eye-level straight-on”;

How virtual_kaf Built This EAT THE PAST English Version Visual — and How to Recreate It

If you’ve ever needed a release announcement visual that feels intimate (not “promo-y”), this is a great reference: a single close-up character, a cool-blue night city blur, and one short line of copy that reads like a lyric. It’s simple on purpose — and that simplicity is exactly what makes it easy to remember, screenshot, and remix.

Why this format travels so well

The first thing you feel here is distance + closeness at the same time: the city is far away (bokeh, haze, blue ambience), but the face is right in front of you. That contrast creates emotional volume without adding narrative clutter. Then the typography does something smart: it doesn’t explain the story — it invites the viewer to supply their own. “Searching for your smile,” is a hook that works whether you’re teasing a song, a reel, a chapter drop, or just a mood.

From a creator-growth angle, this is also a “fast-read” asset. In the first half-second you can parse: who (one subject), where (night city), tone (cool + wistful), and message (one line of text). The background blur is doing the heavy lifting — it makes the image feel cinematic while keeping the attention locked on the face.

Signal Table

Signal	Evidence (from this image)	Mechanism	Replication Action
One-face dominance	Subject fills most of the frame; eye-level gaze	Instant emotional connection; viewers stop because it “looks back”	Lock a medium close-up, centered framing; keep the face unobstructed
Blue-night mood	Cool ambient grade + cyan rim light; soft bloom	Makes the piece feel cinematic and nostalgic without needing plot	Push cool ambient lighting + add a cyan rim; lift blacks slightly
Lyric-like micro copy	“Searching for your smile,” in white serif type	Creates a “caption inside the image” that’s screenshot-friendly	Use a 4–6 word line; keep it legible; reserve negative space for text

Best-fit use cases (and how to adapt)

Music release / teaser cover — keep the close-up + copy; swap the line to a chorus fragment or title tag.
Trailer thumbnail for a series — keep the night bokeh; change wardrobe colors to match your brand palette.
“I’m back” / return post — keep the stare and the soft mouth; change the background to your city or a landmark blur.
Community milestone — keep the framing; change the text to “thank you for…” and add a tiny date stamp (subtle).

Not ideal for

Complex tutorials — you don’t have room for multiple ideas; this format is built for one emotion.
Multi-product promos — adding too many objects breaks the clarity that makes it shareable.

Transfers (3 recipes)

Transfer 1: same mood, different hero
- Keep: cool blue night grade, cyan rim, shallow depth of field
- Change: character design (hair/eyes), wardrobe accent color
- Slot template: {hero} close-up, {city} night bokeh, cool blue grade, cyan rim light, serif lyric line: “{hook}”
Transfer 2: same typography, new scene
- Keep: centered portrait, serif copy layout, background blur
- Change: environment (subway, rooftop, convenience store, rain-lit street)
- Slot template: centered portrait, {scene} at night, strong bokeh, serif text layout, “{line}”
Transfer 3: same face, different emotion
- Keep: lens feel (telephoto), rim light, color palette
- Change: micro-expression + text intention (hopeful / regret / determination)
- Slot template: {hero} medium close-up, cool city night blur, subtle bloom, expression: {emotion}, text: “{promise}”

Aesthetic read: what you can actually recreate

What makes this look “expensive” isn’t a complicated scene — it’s three controllable choices: (1) a background that’s intentionally unreadable (heavy blur), (2) a cool ambient grade with a neon edge, and (3) a face that stays soft and centered. Notice how the subject’s hair picks up a cyan outline: that tiny rim is enough to separate the character from the city lights. The eyes do the opposite: they’re the warmest thing in the frame, which pulls attention straight to the gaze.

Observed → Recreate

Observed	How to recreate (prompt/control)
Soft frontal fill, minimal shadows	“soft front fill, low contrast lighting, minimal hard shadows”
Cyan rim outlining hair/shoulders	“cyan rim light, neon edge glow, cool blue rim on camera-right”
Night city bokeh + haze	“night urban background, heavy bokeh, atmospheric haze, filmic bloom”
Centered close-up with text-safe space	“centered medium close-up, reserve negative space top-left for serif text”

Prompt technique breakdown (think in knobs)

Prompt chunk	What it controls	Swap ideas (EN, 2–3 options)
`subject close-up, calm vulnerable expression`	Stop power and emotional tone	“gentle smile”, “teary eyes”, “determined stare”
`night urban bokeh, atmospheric haze`	Background energy without clutter	“subway platform blur”, “rainy neon street”, “rooftop skyline”
`cool ambient + cyan rim light`	Separation and cinematic feel	“warm tungsten rim”, “magenta rim”, “moonlight rim”
`85mm telephoto, shallow depth of field`	Compression + subject dominance	“50mm natural”, “105mm dreamy compression”, “35mm more environment”
`white serif lyric text layout`	Memorability + screenshot value	“all-caps sans”, “handwritten note”, “typewriter mono”

Quick prompt starter (template)

centered anime 3D portrait, calm gaze, night city bokeh, cool blue ambient, cyan rim light,
85mm telephoto, shallow depth of field, subtle bloom,
white serif text: “Searching for your smile,” with text-safe space top-left

Remix steps (how to converge fast)

Baseline lock (lock these first)

Composition: vertical 9:16, centered medium close-up, face unobstructed
Lighting direction: soft front fill + cyan rim edge
Lens feel: telephoto + shallow depth of field for heavy background blur

One-change rule

Change only 1–2 knobs per run. If you change hair, scene, lighting, and font all at once, you won’t know what actually improved the image.

Example 4-step iteration sequence

Run 1 (base): lock framing + lighting + DOF; keep background as generic “night city bokeh”.
Run 2: adjust rim intensity (“stronger cyan rim”) and bloom (“slightly more glow”).
Run 3: refine typography placement (reserve top-left space; keep text off the face).
Run 4: swap one story element (new hook line or new city) while keeping everything else fixed.