Samuel Ferrero.

Me peleo con modelos de IA, te muestro los resultados y te regalo los promtps.

Yo que me sigas e interactues si te mueve lo que hago | Pollo AI CPP

[Samuel Ferrero. @MiMundoConIA]

con la ia puedes convertir cualquier personaje de ficción en un cartel de cine art decó de los años 20.

peaky blinders lleva años en mi cabeza. tommy, arthur, polly, alfie.

esto es lo que salió. 👇

[Samuel Ferrero. @MiMundoConIA]

Cognee transforma datos no estructurados en un grafo de conocimiento persistente.

plugins para OpenClaw, Claude Code y Hermes Agent.
graduado del GitHub Secure Open Source Program.
API REST + SDK en Python/JS.

repo: https://t.co/7ejh2vFncj

[Samuel Ferrero. @MiMundoConIA]

alguien construyó una memoria permanente para agentes de IA en 6 líneas de código.

se llama Cognee.

92.5% de precisión en recall.
reemplaza RAG tradicional.
conecta datos por significado y relaciones.

esto es lo que hace 👇

[Samuel Ferrero. @MiMundoConIA]

si quieres estar al día de herramientas y frameworks que comparto antes de que se viralicen, únete al telegram.

https://t.co/Gq0nnvVsgd

te dejo el repo en el siguiente post 👇

[Samuel Ferrero. @MiMundoConIA]

esto es arena de playa.

llevas toda la vida pisándola sin saber que tiene esto dentro.

el bloque de estilo macro funciona con cualquier cosa. 👇

[Samuel Ferrero. @MiMundoConIA]

@fiction_log Lo he intentado y no funciona para mi. Seedance 2.0 me da error todas las veces y no me genera el video.

Quiza escogi 2 personajes demasiado famosos y saltan las politicas de uso...

[Samuel Ferrero. @MiMundoConIA]

esto es arena de playa.

llevas toda la vida pisándola sin saber que tiene esto dentro.

el bloque de estilo macro funciona con cualquier cosa. 👇

[Samuel Ferrero. @MiMundoConIA]

incluir un bloque de estilo al prompt cambia el resultado por completo.

la imagen 1 está hecha sin bloque de estilo.

la imagen 2 con el bloque de estilo aplicado.

¿cuál preferirías para una campaña de producto?

te dejo el bloque de estilo en el último post 👇 guárdatelo y pruébalo.

un bloque de estilo es un conjunto de instrucciones técnicas que le dices a la IA para que la imagen parezca una foto real.

este simula las imperfecciones de una cámara real: grano de película, textura de piel, bokeh orgánico, aberración cromática.

el resultado deja de parecer IA y empieza a parecer fotografía editorial.

el truco está en dejar de darle palabras mágicas a la IA y hablarle como un director de cine lo haría con su equipo.

[Samuel Ferrero. @MiMundoConIA]

Prompt:

```
You are a creative director at a top-tier sports advertising agency.
Your task is a two-phase operation: first strategic analysis, then execution.
Do not skip phase one. Do not output phase one.

The three input variables are:
- BRAND: [MARCA]
- PRODUCT: [PRODUCTO]
- COMPOSITION / REFERENCES: [COMPOSICIÓN/REFERENCIAS]

GOVERNING PRINCIPLE — VISUAL TENSION:
Every decision in this poster serves one goal: to generate a question in the
viewer's brain before they can consciously process the image. That question
is what stops the scroll. Not beauty. Not clarity. Tension.

The poster should feel like a photograph taken at exactly the wrong moment —
or exactly the right one. Something is slightly off. Something is more than
it should be. Something is missing where there should be something.
If the image is immediately and completely legible, it has failed.

---

PHASE ONE — STRATEGIC BRIEF (silent — do not output this)

1. TENSION TYPE — derive from the product and composition input.
If [COMPOSICIÓN/REFERENCIAS] names a specific tension type, use it.
If empty, select the tension that best serves this specific product:

a) UNEXPECTED SCALE — the product is too large or too small for its context.
The shoe occupies the entire frame floor to ceiling, making the background
architecture look like a miniature. Or: the shoe is tiny, placed alone on
a vast empty surface that dwarfs it. Scale distortion creates unease.
Best for: footwear, balls, equipment with strong silhouettes.

b) TEXTURE CONTRAST — soft against rough, organic against industrial.
Running shoe fabric pressed against raw concrete. Technical mesh against
worn leather. The materials should not coexist — their meeting creates
the tension.
Best for: footwear, apparel, accessories — anything with distinctive material.

c) IMPOSSIBLE LIGHT — a light source that should not exist in that environment.
A single shaft of warm amber studio light falling on a shoe in a dark
alley. A neon gel light illuminating mesh fabric in an abandoned warehouse.
The light is wrong for the space — which means the eye cannot resolve it.
Best for: any product, especially effective with dark backgrounds.

d) FRAGMENTATION — the product cut by the frame, only a portion visible.
The heel and a fragment of the sole visible at the right edge of the frame.
No full product shot. The brain wants to complete what's missing.
The cropping is aggressive — not accidental, deliberate.
Best for: iconic silhouettes, products with instantly recognizable details.

e) WRONG CONTEXT — a running shoe in a luxury environment, or vice versa.
Trail shoe on a Carrara marble floor. Basketball sneaker in a museum.
The mismatch activates status anxiety, aspiration, or humor simultaneously.
Best for: premium product launches, lifestyle crossover campaigns.

2. ACCENT COLOR — the unexpected choice for this specific brand:

The accent color should feel wrong at first glance, then inevitable.
Standard brand colors are for logos and retail. Campaign colors are for stopping.

Derive the accent based on the brand's established palette and its opposite:
- Nike (typically black/white/red): electric green, acid yellow, or deep violet
- Adidas (typically black/white/blue): volcanic orange, magenta, or burnt sienna
- New Balance (typically gray/navy): saturated red, electric teal, or warm gold
- Asics (typically blue/orange): deep emerald, electric yellow, or cold silver
- Puma (typically black/gold): cobalt blue, neon lime, or blood red
- Under Armour (typically black/red): cyan, deep amber, or bone white
- Unknown/other brand: select the color directly opposite their primary in the
color wheel, pushed to 100% saturation. That is the accent.

This accent appears in ONE element only:
- The single specular highlight on the product
- One thin edge of light on a product detail
- A narrow band of color in an extreme corner of the composition
- A material element of the product (a colorway panel, a sole stripe, a lace)

Everywhere else in the image: desaturated. Black, near-black, white, off-white.
The tension between maximum saturation in one point and zero saturation
everywhere else is what makes the eye unable to look away.

3. TYPOGRAPHY DECISION — binary choice, no middle ground:

Option A — DOMINANT: the brand name or tagline occupies 35–45% of the
poster's total area. Letters so large they become abstract geometric shapes
at close range. The type is integrated in the composition — not floating above
the image but existing in the same physical space as the product.
At this scale, the typography IS the composition.
Use when: the product is partially off-frame (fragmentation tension), or when
the product is small against a large negative space.

Option B — GHOST: the brand name appears in one corner at approximately
8–10% of the poster height. Almost invisible. The viewer must look for it.
Finding it is a small reward that extends the time spent with the poster.
Use when: the product is the full visual protagonist (scale tension, impossible
light), or when the product is so iconic the brand name is already implied.

RULE: never both at the same scale. If headline exists, no tagline.
If tagline exists, no headline at the same size. One message. One size.
Typography that is centered and symmetrical is retail signage, not campaign art.

4. LIGHTING SETUP — one source, oblique, with character:

Determine the single light source direction from the composition and tension type:
- For IMPOSSIBLE LIGHT: the source position is counter-intuitive — a shaft of
light falling from below, or from a direction where no natural source exists.
This is the tension. Commit to the impossibility.
- For other tensions: the light comes from one side, at approximately 30–45°
from the product surface, creating a shadow zone that occupies at least
40% of the product face. The shadow defines the form more precisely than
the color. Do not correct it. Do not add fill.

Light color: colored gel or neon-adjacent — the accent color as the light source,
or its complement. Never clean white studio light. Light that has passed through
something: dust, smoke, a gel, a slit in a wall. Light with a reason.

Specular highlights: exactly placed on one material element of the product.
The gloss of a rubber sole. The weave of a technical mesh. The texture of a
leather panel. One tight, sharp specular that reveals the material at the level
of a forensic photograph. This is the close-up detail that makes the image
feel like real photography, not product render.

5. COMPOSITION RULES — derive the exact layout:

a) BROKEN RULE OF THIRDS: the product is placed at or beyond the edge zones —
at the very edge of the frame, partially cut, or in the extreme corner.
Nothing at the center. The center of this poster is empty.

b) ONE ELEMENT EXITING THE FRAME: the product, a shadow, a light shaft, or
a type element crosses the frame edge. This creates implicit depth and
movement without needing 3D effects. The image feels larger than its frame.

c) DOMINANT DIAGONAL: the primary visual line of the composition is diagonal.
It can be: the product's primary axis, a shadow edge, the light shaft,
the perspective of a surface, the type baseline. One strong diagonal.
It should be obvious even at thumbnail size.

d) NEGATIVE SPACE: 60–70% of the poster is empty — the background color,
unoccupied, providing the visual silence that gives the active zone its weight.
This is not wasted space. It is the pressure that makes the subject exist.

6. WHAT THIS POSTER NEVER CONTAINS — verify each:
- No gradient in the background (solid black or solid off-white only)
- No multiple products in the same plane
- No centered, symmetric typography
- No clean studio lighting without character (no three-point fill)
- No second accent color — the one accent is the only color
- No drop shadow beneath the product (this is the e-commerce watermark)
- No perfect, pristine product condition — the product has texture, use, history.
Not damage. Not wear for its own sake. But the subtle indicators that this
object exists in the real world and has been used by a real person.

---

PHASE TWO — OUTPUT

Generate ONE complete image generation prompt using everything derived above.
Output ONLY the prompt — no preamble, no explanation, no phase labels.

---

Scroll-stopping sports brand campaign poster — [BRAND] — [PRODUCT]

GOVERNING PRINCIPLE:
This image must generate a question in the viewer's brain before they can
consciously process what they are seeing. That pre-conscious tension is the
entire purpose of every decision in this composition.
It is not a product photograph. It is not a lifestyle image.
It is a campaign poster designed to stop a scroll at 0.3 seconds.

COMPOSITION:
[TENSION TYPE from Phase One — full description of how this specific tension
is applied to this specific product in this specific visual context]

The product is NOT centered. It occupies [SPECIFIC POSITION from Phase One —
edge, corner, partially off-frame]. The exact framing:
[Description of cropping, what is visible, what is cut by the frame edge]

One element exits the frame: [SPECIFIC ELEMENT — which part of the product,
which shadow, which light shaft, crosses which edge of the frame].

The dominant visual line of the composition is diagonal:
[Description of the diagonal — what creates it, direction, intensity]

Negative space: [PERCENTAGE]% of the poster is the background color —
[BLACK or OFF-WHITE from Phase One] — empty, unoccupied, with full weight.
No objects, no texture overlay, no gradient intrusion into this zone.

PRODUCT — [PRODUCT NAME AND DESCRIPTION]:
[Full product description translated into visual and material terms —
what the product looks like at the level a photographer would brief a retoucher:
specific materials, their surfaces, the exact colors of each panel or element,
any distinctive design features that make this specific product recognizable.
If the user provides a reference image, the product is reproduced from that image.]

The product is NOT pristine. It is not an e-commerce hero shot.
[Specific material indicators of real use — texture of the sole, slight surface
variation in the upper, the natural behavior of the lace under tension.
Not damage. The signs that this object was made for a body and a body used it.]

LIGHTING:
Single light source — [DIRECTION] — [LIGHT TYPE AND COLOR from Phase One].
[Description of exactly where the light hits the product, what zone falls into
shadow, and why this light source should not exist in this environment if
using impossible light tension]

The shadow occupies approximately 40% of the product face. It is not corrected.
Not filled. The shadow defines the form more precisely than the color.

Specular highlight: [EXACT LOCATION on the product — specific material element].
Tight, sharp, revealing the material texture at forensic resolution.
[Color of the specular — the accent color or a neutral depending on Phase One]

COLOR PALETTE — ABSOLUTE:
Three values only. No exceptions.
- BACKGROUND: [BLACK or OFF-WHITE from Phase One] — solid, no gradient
- DESATURATED BODY: the product in shadow and mid-tone zones is desaturated —
colors present but pulled toward gray, as if seen in low light
- ACCENT — [ACCENT COLOR from Phase One, with hex]: [SINGLE ELEMENT] only.
Maximum saturation. Everything else in the image is at zero saturation relative
to this point. The eye cannot avoid this color because nothing else competes with it.

TYPOGRAPHY:
[TYPOGRAPHY DECISION from Phase One — Option A DOMINANT or Option B GHOST]

[If DOMINANT]:
"[BRAND NAME or TAGLINE]" — [FONT DESCRIPTION: condensed sans-serif, weight,
all caps] — occupying 35–45% of the poster area. Letters large enough to become
abstract geometric shapes at close range. The type exists in the same visual plane
as the product — integrated in the composition, not floating as a layer above it.
Baseline: [DIAGONAL or matching the composition's dominant diagonal].
Position: [SPECIFIC POSITION that reinforces the composition].
Color: [BACKGROUND COLOR INVERTED — if background is black, type is off-white, and vice versa].
No tagline. No brand descriptor. No website URL.

[If GHOST]:
"[BRAND NAME]" — [FONT DESCRIPTION: condensed or geometric sans-serif] —
[CORNER: lower right / upper left] — approximately 8% of poster height.
Color: [SLIGHTLY off background — barely legible, almost invisible].
Nothing else. No tagline. No CTA. The brand name is the only text in the image.

CONTROLLED IMPERFECTION — MANDATORY:
This image must not look like a digital render or AI generation.
It must look like it was taken by a real photographer briefed by a creative director:
- Film grain structure present across the entire image — most visible in the shadow
zones and in the solid background color
- Slight chromatic aberration at the highest-contrast edges — barely perceptible
red-cyan color fringing at the extreme corners
- The specular highlight on the product clips slightly — loses detail at its peak,
as real highlights do in real photography
- Focus breathing: the transition to out-of-focus zones is organic and uneven —
not a uniform blur, but the depth-of-field behavior of a real lens
- The product material surfaces show micro-texture visible at close range —
the grain of rubber, the weave count of mesh, the surface finish of leather —
that makes this image feel like forensic photography of a real object

NEGATIVE SPACE:
The [BLACK / OFF-WHITE] background zone is clean. No texture overlay, no vignette,
no subtle gradient. It is the weight that gives the subject its presence.
Do not fill it. Do not decorate it. Its emptiness is the composition.

WHAT THIS IMAGE EXPLICITLY REJECTS:
- Gradient background of any kind
- Multiple products or elements sharing the same visual plane
- Centered, symmetric composition
- Three-point studio fill lighting
- More than one accent color
- Drop shadow beneath the product
- Pristine, untouched product surface with no material history
- Lifestyle context (no human subject, no action shot, no environment narrative)
- Typography at a medium, unremarkable size

Photorealistic. Sports campaign quality. Aspect ratio 4:5 vertical.
The image should feel like a campaign tear-sheet from Dazed, Highsnobiety, or
a marquee outdoor campaign — real, specific, and impossible to scroll past.
No watermark. No production credit. No network logo.
```

[Samuel Ferrero. @MiMundoConIA]

una sesión de fotos de producto para nike cuesta 3.000€ mínimo.

estos cuatro carteles los generé con un prompt en 10 minutos.

nike, adidas, new balance, puma. cuatro marcas, cuatro estilos. te dejo el prompt 👇

[Samuel Ferrero. @MiMundoConIA]

cada vez que subo algo nuevo, lo aviso en el telegram. si no quieres perdértelo 👇

https://t.co/QEkRowL4Y5

el prompt, en el siguiente post 👇

[Samuel Ferrero. @MiMundoConIA]

el bloque base y las 4 variantes 👇

**Cuándo usar:** insectos, flores, texturas naturales, minerales, gotas de agua, piel, cualquier sujeto donde el detalle microscópico es el protagonista. Elimina el look limpio y artificial de IA. El resultado debe pasar por una fotografía macro de revista científica o portfolio profesional de naturaleza.

```
Ultra-realistic macro photography — mandatory:
Shot on Canon MP-E 65mm f/2.8 1-5x Macro lens, f/8, ISO 200, 1/200s flash sync.
Extreme shallow depth of field — razor-thin focal plane, progressive bokeh falloff.
Ring flash diffused, catchlights visible on subject surface.
Subject fills 80% of frame. Microstructure fully resolved: cellular texture,
surface hydration, microscopic hairs, dust particles, iridescent light scatter.
True-to-life color grading — no oversaturation. Natural specular highlights.
RAW-processed, no HDR artifacts. Film grain at 12.8%. 8K resolution output.
```

**Variante — Insecto:**
```
[Bloque macro base arriba] +
compound eye facets fully resolved, exoskeleton micropattern and chitin iridescence,
wing venation translucency with backlight diffusion, leg joint micro-articulation,
antenna filament catch-light, setae (microscopic hairs) individually rendered.
```

**Variante — Flor:**
```
[Bloque macro base arriba] +
petal cell structure and epidermal wax layer, pollen grain distribution on stamen,
nectar bead surface tension and meniscus, pistil micro-texture, vein reticulation.
```

**Variante — Agua / Gota:**
```
[Bloque macro base arriba] +
droplet meniscus geometry, refraction caustics on surface below, suspended micro-particles,
contact angle precise, internal reflection of environment distorted through sphere.
```

**Variante — Piel / Textura:**
```
[Bloque macro base arriba] +
pore microrelief and sebum sheen, fine vellus hair catch-light, dermal ridge pattern,
micro-cracks in dry zones, subcutaneous color variation visible through thin skin.
```

**Variante — Mineral / Cristal:**
```
[Bloque macro base arriba] +
cleavage plane reflectance and step geometry, inclusion depth and color,
birefringence fringe at polarized edges, crystal face micro-etch patterns.
```

**Motion prompt (para animar con HappyHorse I2V):**
```
Microscopic camera drift — imperceptible push-in at 0.3x speed, focal plane
holding on primary subject. Ambient breath: micro-vibration from air current,
surface tension quiver on liquid droplets, hair filaments oscillating subtly.
Lighting stable — ring flash held, specular catchlights locked. Background bokeh
spheres drift laterally at 10% opacity shift. Identity-stable subject geometry.
Single primary motion only: subject tremor. No camera shake.
```

**Negative prompt recomendado:**
```
no oversaturation, no HDR glow, no plastic sheen, no digital noise,
no motion blur on subject, no lens flare artifacts, no AI smoothing,
no depth map artifacts, no vignette crush

[Samuel Ferrero. @MiMundoConIA]

si no quieres perder el prompt, en el telegram los tienes todos listos para copiar.

https://t.co/QEkRowL4Y5

las variantes abajo 👇

[Samuel Ferrero. @MiMundoConIA]

una mantis religiosa. una flor de loto. una gota de agua. tela vaquera.

le metí el mismo bloque de estilo macro a los cuatro.

el bloque completo con las 4 variantes, en el siguiente post 👇

[Samuel Ferrero. @MiMundoConIA]

la primera imagen ya no parecía IA.

le había metido un prompt de 400 palabras. lente, apertura, comportamiento de la luz, textura de las escamas. el del ala de mariposa de ayer.

dos iteraciones más y no había forma de distinguirlo de una foto real.

llevo semanas con la misma herramienta y prompts de dos líneas sin llegar ni cerca de esto.

no es la herramienta. es el nivel de detalle con el que le hablas.

[Samuel Ferrero. @MiMundoConIA]

Prompt:

```
You are an anime production designer working in the tradition of high-production
television anime of the 2020s. Your task is a two-phase operation: first research,
then generate. Do not skip phase one. Do not output phase one.

The three input variables are:
- CHARACTER: [PERSONAJE]
- FRANCHISE: [FRANQUICIA]
- COMPOSITION / REFERENCES: [COMPOSICIÓN/REFERENCIAS]

CRITICAL DISTINCTION — READ FIRST:
This image is an ANIME CHARACTER SHEET (settei / character design sheet) —
the kind produced by professional anime studios for the animation team to
reference while drawing a character across hundreds of scenes.
NOT a poster. NOT an action scene. NOT a background illustration.
NOT a photorealistic render. NOT a video game screenshot.
NOT chibi or super-deformed proportions.

The specific production context: this sheet would be produced by a studio like
Ufotable, Madhouse, or MAPPA for their own animation team. It is functional design
documentation — precise, clean, informative, beautiful. Every element in the sheet
serves a design or production purpose.

---

PHASE ONE — CHARACTER DESIGN BRIEF (silent — do not output this)

1. ANIME TRANSLATION — the character as a high-production anime face:

The Ufotable-style anime aesthetic translates ANY character from any medium
into anime design language — not by making them look Japanese, but by applying
the visual grammar of contemporary anime character design:
clean line hierarchy, simplified but expressive features, color-coded identity.

Analyze this character and determine:

a) FACE STRUCTURE IN ANIME GRAMMAR:
- Strong geometric features (jaw, cheekbones) → retain the structural quality
in simplified form. The jaw angle stays, the detail reduces.
- Softer, rounder features → slightly larger eyes relative to face,
more visible emotional range in default expression.
- Aged or weathered face → retain age markers (lines, eye shape, brow weight)
but in clean line form — not photorealistic wrinkles, but designed age.
- Unconventional features → find the single most recognizable element
and amplify it in design language. This is the character's visual signature.

b) EYE DESIGN — the most important character design decision:
In contemporary anime production, eye design is the first thing animation
teams memorize. It must communicate the character's narrative archetype:
- HERO / PROTAGONIST: bright, visible iris with clear specular highlights,
medium-large relative to face. The "alive" quality of the eye carries the
character's will and determination. Demon Slayer / Fate protagonist grammar.
- VILLAIN / ANTAGONIST: narrower eyes, heavier upper lid line, cool or
unusual iris color. The specular highlight is smaller or positioned unusually.
The eye communicates controlled intensity, not warmth.
- MORALLY COMPLEX / ANTIHERO: asymmetric visual weight — one eye carries
more shadow or detail than the other, or the iris color has an unusual
depth suggesting layers. The eye holds something back.
- SUPPORTING / EXPRESSIVE: eyes that read quickly at small size — clear,
readable iris, expressive brow. Designed for comedic or emotional beats.

c) CANONICAL EXPRESSION SET — determine four expressions for this character:
Always includes NEUTRAL (their resting face — how they look between scenes).
Then choose THREE from: FOCUSED/DETERMINED, ANGRY/INTENSE, SMILE/WARM,
COLD/CALCULATING, SURPRISED/SHOCKED, PAINED/CONFLICTED, SMUG/CONFIDENT.
Choose based on which expressions this character actually uses in their story.
Walter White ≠ Jesse Pinkman ≠ Gus Fring even from the same franchise.

d) SIGNATURE ACCESSORY OR WEAPON:
The one physical object most associated with this character. If they have
a weapon: describe it in terms of design (not just "sword" but the specific
proportions, any distinctive features, the handle wrapping, the blade shape).
If they have no weapon: their most iconic non-weapon accessory (the hat,
the glasses, the coat, the phone).
This object gets a dedicated detailed view in the sheet at consistent
line weight with the character.

2. PALETTE EXTRACTION — the character's design colors:

Identify the primary palette of this character as it would appear in anime
production. Anime characters have a defined, limited color set that functions
as their identity marker:

a) SKIN TONE: exact description in anime terms — the base tone, the shadow
tone (typically a slightly cooler, darker variation), the highlight tone
(a lighter, slightly warmer variation for specular). No more than three
skin values in standard cel-shading.

b) HAIR: base color + shadow color + optional highlight color.
Anime hair is stylized toward a simplified, striking version of the
real color — black hair has deep indigo or dark blue shadows,
blonde hair has warm ochre shadows, white/silver hair has cool lavender shadows.
Choose what makes this character's hair most distinctively theirs.

c) MAIN COSTUME: 2–4 colors for the primary outfit. These should derive
from the character's actual costume — even if this is a real-world
character adapted to anime, their clothing should feel consistent with
their established visual identity. A chef in whites → specific off-white +
shadow gray. A dark suit character → charcoal + black + one accent.

d) ACCENT COLOR: the signature color, often in small quantities, that reads
as "this character" at a glance. Tanjiro's earrings. Geralt's witcher
medallion. The single color that appears in eyes, a detail, or an accessory
and anchors the whole palette. This color appears in the palette swatches
with a slightly larger swatch to signal its identity function.

e) PALETTE SWATCH ROW: 6–8 swatches arranged horizontally, labeled
below each swatch in romaji (skin / shadow / hair / hair-shadow /
costume-1 / costume-2 / accent / eye). Swatches are solid color rectangles,
approximately 1:1.5 ratio, evenly spaced.

3. JAPANESE NAME TRANSLATION:
Translate the character's name to Japanese (katakana for foreign names,
kanji/hiragana for names that have established Japanese forms).
Include romaji transliteration below the Japanese.
Examples:
- Walter White → ウォルター・ホワイト / Wōrutā Howaito
- Geralt de Rivia → ゲラルト・デ・リヴィア / Geralt de Rivia
- Tanjiro Kamado → 竈門炭治郎 / Kamado Tanjirō
- Tony Stark → トニー・スターク / Tonī Sutāku

---

PHASE TWO — OUTPUT

Generate ONE complete image generation prompt using everything derived above.
Output ONLY the prompt — no preamble, no explanation, no phase labels.

---

Anime character design sheet (settei) — [CHARACTER NAME], [FRANCHISE]

STYLE AND MEDIUM — MANDATORY:
This image is a professional anime character design reference sheet in the visual
tradition of contemporary high-production anime studios (Ufotable, Madhouse, MAPPA).
The style reference is specifically Ufotable's character design approach as seen
in Demon Slayer: Kimetsu no Yaiba and Fate/stay night: Unlimited Blade Works:
clean, precise ink-like outlines with consistent line weight, anime-accurate
proportions, and cel-shading with clearly separated shadow and light zones.

NOT a poster. NOT an action scene. NOT photorealistic. NOT chibi or super-deformed.
NOT western animation style. NOT a video game render.
This is production documentation — functional, clear, and visually precise.

BACKGROUND:
Pure white or very light neutral gray (#F4F4F4 equivalent) — completely clean.
No gradients, no environment, no atmospheric elements. The sheet itself is
the visual content. Grid lines or registration marks are optional and subtle.

---

SHEET LAYOUT — THREE VIEWS (TOP SECTION):
The upper three-quarters of the image is divided into three panels side-by-side:

LEFT PANEL — FRONT VIEW:
Full-body front view, feet to top of head visible, centered in the panel.
Proportion reference: standard anime proportions, approximately 7–7.5 heads tall
(not super-deformed, not realistic). The figure faces directly forward,
arms relaxed at sides or in a characteristic idle stance.
Scale tick marks or height reference lines can appear at the left edge of this panel.
This is the primary design view — all costume details fully visible, coloring complete.

CENTER PANEL — SIDE PROFILE (3/4 or full lateral):
Full-body side view (RIGHT side facing LEFT in the image, standard animation convention).
Same scale as the front view — the horizontal baseline of the feet aligns
across all three panels.
This view reveals: the depth of the hair design, the profile line of the face,
the silhouette of the costume from the side.

RIGHT PANEL — BACK VIEW:
Full-body back view, same scale and baseline alignment.
Reveals: back of hair (a distinct design element in anime), back of costume,
any accessories or details not visible from front or side.

The three figures are drawn at exactly the same scale, with baseline alignment
so their heights match horizontally — this is the functional requirement of a real settei.
A subtle horizontal baseline rule runs across all three panels at foot level.

---

EXPRESSION SHEET (MIDDLE SECTION):
A horizontal row of four bust portraits (head and upper chest), evenly spaced,
each in a separate implied cell. These show the character's face only —
consistent hairstyle, no costume variation.

Expression 1 — NEUTRAL:
The character's resting expression — how they appear between emotional beats.
[NEUTRAL EXPRESSION DESCRIPTION from Phase One: specific configuration of brow
position, eye openness, and lip line that defines this character at rest]

Expression 2 — [SECOND EXPRESSION from Phase One]:
[Full description of brow configuration, eye shape change, mouth position]

Expression 3 — [THIRD EXPRESSION from Phase One]:
[Full description of brow configuration, eye shape change, mouth position]

Expression 4 — [FOURTH EXPRESSION from Phase One]:
[Full description of brow configuration, eye shape change, mouth position]

Each expression portrait is the same scale. The same line art is used for the
base face structure — only the brow position, eye shape, and mouth line change
between expressions. This is intentional: it demonstrates design consistency
across emotional states, as a real animation studio would require.

---

ACCESSORY / WEAPON DETAIL (LOWER LEFT):
A standalone detail view of [SIGNATURE ACCESSORY/WEAPON from Phase One],
drawn at a larger scale than it appears on the character (typically 150–200% scale),
from the angle that best reveals its design. This view:
- Shows all the design elements: shape, any decorative details, material indication
through line work rather than texture
- Is drawn with the same clean line art style as the character
- Has no background — the object floats in the white space
- May include 1–2 additional angle views if the design is complex
A small label in the style of Japanese production notes (brief, in a light gray font)
identifies the object: "[OBJECT NAME]" in romaji + optional Japanese

---

PALETTE SWATCHES (LOWER RIGHT):
[PALETTE SWATCH ROW from Phase One — 6–8 swatches]
A horizontal row of solid color rectangles, each approximately square with a
slight height elongation. Below each swatch: the label in small, clean sans-serif
(the type that appears in real production documents). The accent color swatch
is very slightly larger than the others.
The swatch row is positioned in the lower right quadrant of the sheet,
with [CHARACTER NAME in Japanese + romaji from Phase One] displayed in larger
text above the swatches.

---

LINE ART STYLE — MANDATORY:
Ufotable-style clean line art:
- Consistent, confident outline weight — not the variable organic weight of manga
or watercolor illustration. The line is precisely weighted: slightly heavier on
outer contours, slightly lighter on interior details.
- Cel-shading with two tones maximum per color zone: base color and one shadow tone.
Shadow edges are crisp, not blended — a clean edge between light and shadow,
as in physical cel animation.
- Hair is drawn as a simplified shape language — individual strand detail is
present but organized into groups of strands moving in the same direction,
not individually fine. The overall hair silhouette is the primary design element.
- Eyes follow the [EYE DESIGN from Phase One]: iris detail, specular highlight
placement, and overall shape consistent across all views and expressions.

---

COLOR SYSTEM — [PALETTE FROM PHASE ONE]:
All color in this sheet is flat cel-shading — no photorealistic lighting,
no ambient occlusion, no soft gradients. Each color zone has exactly two values:
base tone and one shadow tone. The shadow is applied consistently as if a single
directional light source is above and slightly to the left of the character.
In the expression portraits and detail views, the same shadow direction applies.
No color in this sheet deviates from the established character palette.

CHARACTER — [CHARACTER NAME]:
[Full description from Phase One:
- Face structure in anime design language — which features are simplified,
which are amplified, what makes this face immediately identifiable at
the line art level
- Eye design: exact style, iris color, highlight position, upper lid weight
- Hair design: silhouette description, color in anime terms, key design details
- Costume description at the level of the three-view sheet — what is visible
from front, what from side, what from back, any notable design details
that need careful rendering
- Idle stance: the position of the hands and body in the front view]

FRANCHISE TITLE TREATMENT:
In the upper corner of the sheet (typically upper right), in small but readable
production-document style text: "[FRANCHISE NAME]" and below it "[CHARACTER NAME]"
in the same clean sans-serif. Below that: [Japanese name + romaji from Phase One].
This text functions as the production identification header of the sheet —
it is not decorative, it is informational. Size: approximately 14–16pt equivalent
at final image resolution.

MANDATORY CONSTRAINTS — WHAT THIS IMAGE NEVER CONTAINS:
- No backgrounds, scenes, or environmental elements of any kind
- No action effects, particles, or speed lines — this is a reference sheet,
not an action frame
- No photorealistic skin texture, hair simulation, or lighting effects —
this is cel-shaded, not rendered
- No text in English mixed with Japanese (use romaji for pronunciation,
Japanese for the name rendering)
- No comic book or western animation proportions — standard contemporary anime
7–7.5 head proportions apply throughout
- No inconsistency in scale between the three views — the baseline alignment
is a functional requirement. Feet are at the same height across all panels.
- No variation in the character's design between the three views — same hair
shape, same costume, same colors. The views must be internally consistent.
- No facial expressions in the three-view section — front, side, and back are
neutral/idle. Expressions are only in the expression row.
- No gradients, soft shadows, or ambient occlusion — strictly flat cel-shading

Aspect ratio: 3:4 vertical — character sheet format.
Resolution: high enough to make the palette labels and Japanese text legible.
No watermark. No production studio logos. No copyright notices.
```

[Samuel Ferrero. @MiMundoConIA]

demon slayer: infinity castle lleva más de 300 días en cines.

todavía sin streaming.

mientras esperamos, le pedí a la IA que generara esto 👇

[Samuel Ferrero. @MiMundoConIA]

cada vez que subo algo nuevo, lo aviso en el telegram. si no quieres perdértelo 👇

https://t.co/QEkRowL4Y5

el prompt, en el siguiente post 👇

[Samuel Ferrero. @MiMundoConIA]

@jucaripo Pues yo tengo pendiente Better Call Saul. Me la apunto 😉

[Samuel Ferrero. @MiMundoConIA]

breaking bad sigue siendo la mejor serie de televisión que he visto.

walter, jesse, gus, saul.

¿y si hubiera sido un manga japonés de los 90?

esto es lo que salió 👇

[Samuel Ferrero. @MiMundoConIA]

Prompt:

```
You are a manga art director working in the tradition of Japanese physical manga
of the 1990s. Your task is a two-phase operation: first research, then generate.
Do not skip phase one. Do not output phase one.

The three input variables are:
- CHARACTER: [PERSONAJE/ELEMENTO]
- FRANCHISE: [FRANQUICIA]
- COMPOSITION / REFERENCES: [COMPOSICIÓN/REFERENCIAS]

CRITICAL DISTINCTION — READ FIRST:
This image is a PHYSICAL MANGA PAGE PANEL — specifically a character poster panel
as it would appear in a collected volume (tankōbon) of a 1990s Japanese manga.
NOT anime. NOT digital illustration. NOT modern webtoon or digital manga.
NOT color. NOT a western comic.

The specific physical medium: black ink on manga paper (slightly yellowed, with
the slight texture of newsprint-adjacent paper), printed using offset lithography
that makes screentone dots visible under magnification, with ink that shows the
actual pressure variation of a G-pen or kabura-pen nib moving across paper.
The image has been scanned from a physical volume — slight paper texture visible
in the white areas, screentone patterns crisp but with the very slight registration
imperfection of physical printing.

---

PHASE ONE — CHARACTER INTELLIGENCE BRIEF (silent — do not output this)

1. PHYSICAL TRANSLATION — the character as a 90s manga face:

The 90s manga aesthetic is NOT about making characters look Japanese.
It is about translating ANY face into the visual language of expressive
ink line work — emphasizing the features that carry the most narrative weight.

Analyze this character's face and determine:
a) FACE STRUCTURE EMPHASIS: What are the defining geometric qualities of
their face that a G-pen would emphasize?
- Sharp jaw and angular features → thick, confident contour lines,
minimal interior detail — the jaw IS the character (Walter White,
Gus Fring type)
- Softer or younger face → more interior line detail, larger eye space,
softer contour weight — the expression is the character (Jesse type)
- Neutral, controlled face → the tension between minimal line detail
and heavy shadow placement — calm surface over internal intensity
(Urasawa Monster villain grammar)
- Expressive, mobile face → dynamic line weight variation, exaggerated
but not cartoony — high storytelling efficiency per line (Saul type)

b) EYE TREATMENT — the most important decision in manga translation:
In 90s seinen/shonen manga, eyes communicate the character's narrative role:
- COLD ANTAGONIST / VILLAIN: eyes partially shadowed by heavy blacks —
the classic Urasawa/Miura technique where a horizontal band of shadow
across the upper face makes the eye whites glow from darkness.
The iris is detailed but surrounded by threat. Walter White, Gus Fring.
- PROTAGONIST IN CRISIS: eyes large relative to the face, visible irises
and pupils, the whites visible below the iris (manga stress marker),
eyebrows angled inward — vulnerability and determination simultaneously.
Jesse Pinkman register.
- CONTROLLED POWER: eyes narrow, pupils precise, no visible whites below
the iris — the gaze of someone who has already decided. Very fine line
for the upper lid, heavy line for the lower shadow.
- EXPRESSIVE / COMIC-DRAMATIC: eyes animated, variable — wide for
surprise/comedy, narrowed for scheming, never fixed in one register
within the image (even a still image suggests the sequence of expressions
this face cycles through). Saul register.

c) SIGNATURE PHYSICAL ELEMENT: What one physical detail, when drawn in
manga ink style, instantly identifies this character?
(The Heisenberg hat shadow / the Gus corporate neutrality of the glasses /
Jesse's hoodie fabric / Saul's wide-lapel suit collar)
This element receives the most careful ink detail in the image.

2. MANGA ARCHETYPE — assign ONE of these, it determines everything:
- COLD ANTAGONIST: the monster behind the human mask — Urasawa grammar.
Restrained expression. Heavy shadows. The horror is in the calm.
- PROTAGONIST IN CRISIS: the hero being broken and rebuilt — Inoue grammar.
Dynamic line, visible suffering, eyes that tell the story before the body.
- VILLAIN WITH VISIBLE SEAMS: controlled power with one hairline crack —
Miura grammar. The heavy black areas press against a single moment of
visible humanity or weakness.
- EXPRESSIVE SECONDARY: the scene-stealer who is never the protagonist —
comedic-dramatic register that appears in every 90s manga for contrast
and relief. Bigger gestural lines. More exaggerated without losing weight.

3. MASTER REFERENCE — from [COMPOSICIÓN/REFERENCIAS] or derived:

IF a specific manga or artist is named:
- BERSERK / MIURA: Maximum line weight contrast. The thickest outlines
in the panel vocabulary. Hatching is dense and directional — not casual.
Screentones are used dramatically: large zones of single-density tone
behind the figure creating a felt sense of atmospheric weight. The face
has almost sculptural ink volume. Drama is the default state.
- MONSTER or 20TH CENTURY BOYS / URASAWA: Efficient, precise line work.
Nothing wasted. The horror is in the normality — the line is calm but
the shadow placement is unsettling. Screentones are subtle, used for
mid-tones and environmental suggestion, not for drama. A single wrong
shadow angle does more work than twenty speed lines.
- VAGABOND / INOUE (brushwork period): Ink wash qualities alongside
precise line. Variable line that shows the brush acceleration — thin
where the brush moves fast, thick where it slows. Extreme confidence.
The line IS the character's energy, not just their outline.
- SLAM DUNK / INOUE (pen period): Cleaner, more structured line than
Vagabond. Dynamic body poses. Screentones used for uniform gray zones.
Expression is exaggerated in the shonen register — bigger, louder,
more immediate. The sports manga grammar: even a still image feels
like a frozen moment of motion.

IF EMPTY — derive from the archetype:
- COLD ANTAGONIST → Urasawa (Monster)
- PROTAGONIST IN CRISIS → Inoue pen period (Slam Dunk)
- VILLAIN WITH VISIBLE SEAMS → Miura (Berserk)
- EXPRESSIVE SECONDARY → Inoue early period or Toriyama adjacent

4. SCREENTONE PLAN — specific decisions per zone:
Screentone is not a uniform texture. Each zone has a specific decision:

a) SHADOW ON FACE: What density? Light shadow (10% dot halftone — subtle),
medium shadow (30–40% dot halftone — present, atmospheric), or heavy
shadow (60%+ dot — almost black, used for dramatic villain framing)?

b) CLOTHING: Does this character's clothing receive screentone or hatching?
- Screentone is for fabric with volume and texture (suits, hoodies)
- Hatching (parallel ink lines) is for harder surfaces or when the
clothing needs to feel physically drawn rather than printed
- Which specific garment zones get which treatment?

c) BACKGROUND: Is the background a screentone field, speed lines, or
environment-specific hatching?
- Screentone field: a single tone providing atmospheric separation
between figure and ground (the most common choice)
- Speed lines: radiating from behind the figure — used for moments
of high psychological intensity, not for action
- Hatching with environmental detail: only when a specific environment
element is narratively important (the lab, the restaurant, the office)

d) MAXIMUM CONTRAST ZONE: where is the purest white and the deepest black
in this image? This contrast pair is the visual anchor of the panel.

5. KATAKANA / ROMAJI NAME:
Transliterate the character's name into katakana.
Examples: Walter White → ウォルター・ホワイト, Jesse → ジェシー,
Gus Fring → グス・フリング, Geralt → ゲラルト
This appears as small text in one corner of the panel frame — as character
identification text appears in actual manga panels.

6. FRAMING ANGLE — from character archetype and master reference:
- COLD ANTAGONIST / VILLAIN: slight counter-angle (camera below eye level,
looking up) — gives authority and threat without action. The character
fills the upper portion of the frame.
- PROTAGONIST IN CRISIS: eye level or slight overhead angle — creates
psychological vulnerability without diminishing the character physically.
- CONTROLLED POWER: exact eye level — the character meets the reader's
gaze on equal terms, which is itself a kind of dominance.
- EXPRESSIVE SECONDARY: variable angle that favors the character's most
expressive quality — usually eye level with the body at a slight turn.

---

PHASE TWO — OUTPUT

Generate ONE complete image generation prompt using everything derived above.
Output ONLY the prompt — no preamble, no explanation, no phase labels.

---

1990s physical manga panel — character portrait — [CHARACTER NAME], [FRANCHISE]

MEDIUM — MANDATORY — physical manga of the 1990s, black ink and screentone:
This image must be rendered as an authentic manga panel from a physical tankōbon
volume of the 1990s Japanese manga tradition — specifically in the visual language
of [MASTER REFERENCE from Phase One: Berserk / Monster / Vagabond / Slam Dunk].
This is NOT anime. NOT color manga. NOT digital illustration. NOT webtoon.
NOT a western comic book. NOT modern manga with clean digital lines.

The physical reality of the medium: this panel was drawn with a G-pen nib
on manga paper, inked by hand, and screentone was applied as physical adhesive
sheets cut and burnished onto the paper. The image was then scanned and offset
printed. Every mark in this image has the physical character of that process.

COLOR PALETTE — ABSOLUTELY FIXED:
Three values only. No exceptions.
- Pure black (Sumi ink): outlines, deepest shadows, solid fills, speed lines
- Pure white (paper ground): the lightest zones, highlight areas, eye whites
- Grey screentone (halftone dot patterns in specific densities): mid-tones,
atmospheric depth, fabric texture, background zones
There is no fourth color. There is no gradient between these three values.
Every tone in the image is either black, white, or a specific screentone density.

LINE WORK — [MASTER REFERENCE] STYLE:
All forms are defined by ink lines drawn with a pen nib — not vectors, not
brush simulation filters, not AI-generated smooth curves. The line has:
- Variable width within a single stroke: thicker where the nib slows or
turns, thinner where it accelerates through a curve
- Slight irregularity — real pen on paper shows micro-pressure variation
that no digital tool perfectly replicates. The outer contour of the face
has these micro-variations. They are not errors — they are proof of hand.
- Weight hierarchy: THICKEST for outer silhouette and major form separations,
MEDIUM for secondary forms (facial features, major folds), FINEST for
internal details (hair strands, fabric texture, screentone boundary lines)

[MASTER-SPECIFIC LINE CHARACTER:
- Miura/Berserk: maximum weight contrast, the thickest outlines in the panel
vocabulary, hatching is dense and has direction — not casual cross-hatching
but parallel lines following the form, like pressure on a surface
- Urasawa/Monster: efficient precision — every line chosen, nothing decorative.
The calm of the line makes the shadow placement feel more unsettling.
- Inoue/Vagabond: the line shows the brush's speed — where it moved fast
(thin, decisive) and where it slowed (thick, weighted). The line IS the energy.
- Inoue/Slam Dunk: cleaner, more structured, dynamic body grammar, expression
in the shonen register — bigger, more immediate, readable from across the room]

CHARACTER — [CHARACTER NAME]:
[Full physical description translated into manga ink visual language from Phase One:
- Face structure in terms of line weight decisions: which features are emphasized
by thicker or more deliberate ink lines
- Eye treatment: exactly which shadow technique is applied, precise description
of the eye style for this character's archetype
- The signature physical element described at the level of ink mark detail
- Expression: the specific emotional state rendered in this manga vocabulary —
not a general mood, but the specific configuration of brow angle, lip line,
and eye shape that produces this expression in ink line language]

[FRAMING ANGLE from Phase One with precise description of what this means
for the panel: where the chin falls, how much of the torso is visible,
the relationship between the face and the panel frame]

SCREENTONE APPLICATION — [SPECIFIC PLAN FROM PHASE ONE]:
Screentone patterns in this image follow the decisions derived from the
character analysis. Each zone has a specific treatment:

FACE SHADOW ZONES: [DENSITY from Phase One — 10%, 30–40%, or 60%+]
halftone dot screentone applied to [SPECIFIC FACIAL ZONES — the underside
of the jaw, the shadow cast by the brow ridge, the hollow of the cheek].
The dots are visible as dots — not blurred into a gray tone. The halftone
screen is approximately 40–60 lines per inch, consistent with physical manga
printing of the era.

CLOTHING: [SCREENTONE or HATCHING, which zones, which density or direction]
[CHARACTER NAME]'s [SPECIFIC GARMENT] receives [SPECIFIC TREATMENT].
The fabric has volume — the screentone or hatching follows the form of the
body beneath it, lighter at the peaks of folds, darker in the valleys.

BACKGROUND: [SPECIFIC BACKGROUND TYPE from Phase One]
[If screentone field: a single [DENSITY] screentone providing atmospheric
separation, vignetting slightly darker at the edges of the panel frame]
[If speed lines: radiating from a point [LOCATION], converging behind
the character — the lines are drawn with a ruler but show slight ink variation]
[If environmental hatching: [SPECIFIC ELEMENTS] rendered in ink line work
at a lower detail level than the character — they recede, the figure advances]

MAXIMUM CONTRAST ZONE:
The purest white in this panel: [SPECIFIC LOCATION — the whites of the eyes /
the collar of the shirt / a specific highlight on the face or object].
The deepest black: [SPECIFIC LOCATION — the shadow band across the eyes /
the solid fill of the hair / the darkest zone of the background].
These two zones are the visual poles of the panel — the eye moves between them.

PANEL FRAME AND TEXT:
A thin black panel border frames the composition — a clean ink line that
defines the panel edge, as every manga panel has. The border weight is
consistent and precise — approximately 2–3pt equivalent.

In [CORNER — upper right / lower left] of the panel frame, in very small
sans-serif or technical text (the kind that appears in actual manga for
character names and chapter information): [KATAKANA NAME from Phase One].
This text is not decorative — it is part of the informational system of
the manga page.

FRANCHISE TITLE TREATMENT:
"[FRANCHISE NAME]" rendered in manga impact lettering — bold, with slight
stroke variation that suggests hand-drawing (not vector perfect).
The lettering may have a thin outline or a drop shadow in screentone.
Position: [TOP of panel in a header band / BOTTOM of panel in a footer band].
The Japanese equivalent or phonetic rendering appears below in smaller text
if the franchise has a natural Japanese form.
If the franchise is already a Japanese property, use the original title.

HATCHING SYSTEM — where screentone is not used:
In zones where screentone would not be physically applied in a real 90s manga
production (metallic surfaces, certain background elements, character details
that need a drawn quality rather than a printed quality), ink hatching is used:
- PARALLEL HATCHING: evenly spaced lines following the form — used on curved
surfaces, indicating light direction
- CROSS-HATCHING: for deeper shadow than screentone can achieve — used where
black is called for but a solid black fill would lose form
- CONTOUR HATCHING: lines following the 3D form of the surface — used on faces
in the Miura tradition, following the bone structure beneath the skin

MANDATORY CONSTRAINTS — WHAT THIS IMAGE NEVER CONTAINS:
- No color of any kind — not even a hint of warmth or coolness in the "black"
or "white". The blacks are ink. The whites are paper. The grays are dots.
- No smooth gradients — the only tonal transitions are: hard edge between black
and white, or the visual averaging of halftone dots at specific densities.
- No digital line perfection — the line has organic variation from nib pressure.
A perfectly uniform digital stroke of even width is WRONG for this style.
- No modern manga aesthetics — no large sparkle effects, no speed-blur post
effects, no color-correction artifacts, nothing that couldn't be produced
by physical pen, ink, and screentone sheet on paper.
- No photorealistic backgrounds — backgrounds are schematic, symbolic, or
atmospheric. They support the figure, they do not depict a real space.
- No neutral expressions on characters whose archetype calls for emotional charge.
Even a contained expression is charged — the neutrality is performed, not empty.
- No more than three values (black, white, screentone gray). If a fourth value
appears, collapse it into the nearest of the three.

Aspect ratio: 2:3 vertical — standard manga panel/poster format.
The image must read as a physical scan of a printed manga page:
paper texture visible in the white areas (very subtle), screentone dots visible
as distinct dots (not blurred), ink lines with the slight organic character
of a real nib on paper.
No watermark. No modern UI elements. No color correction artifacts.
```

[Samuel Ferrero. @MiMundoConIA]

cada vez que subo algo nuevo, lo aviso en el telegram. si no quieres perdértelo 👇

https://t.co/QEkRowL4Y5

el prompt, en el siguiente post 👇