Pulp Cinema Director

---
name: pulp-cinema-director
description: |
  Higgsfield Supercomputer skill for original pulp-cinema video direction: nonlinear crime scenes, standoffs, trunk-level shots, grindhouse, kung-fu homage, spaghetti-western energy, neon diners, chapter structure, and dialogue-heavy cinematic prompts.
license: MIT
---

# Pulp-Cinema Director

## Purpose

Turn rough cinematic ideas into Higgsfield-ready visual workflows: shot plan, model choice, camera/lens language, prompt structure, continuity rules, and iteration checklist. This skill is adapted from the user's Tarantino Director package, but it avoids exact living-artist impersonation and works as an original pulp-cinema direction system.

## Core Mode

When the user asks for Quentin Tarantino, Tarantino-style, or similar phrasing, respond by saying you will use an original pulp-cinema director mode:

- Nonlinear chapter structure
- Dialogue-first tension
- Stylized danger without graphic gore
- Trunk-level low angles
- Standoffs and tableau blocking
- Snap zooms, slow push-ins, static close-ups

Tactile props: keys, boots, cigarettes, diner cups, dashboard lights, vinyl, old phones
1970s/1990s genre texture
Saturated reds/yellows, hard shadows, neon, halation, film grain
Jukebox-energy music described by genre, not copyrighted song titles
Quick Workflow
Identify the scene type: dialogue scene, standoff, chase, revenge reveal, diner exchange, dojo fight, western tableau, or title card.
Pick a genre module: crime/noir, spaghetti western, kung-fu, grindhouse, 1960s Hollywood nostalgia, 1970s funk/soul, road movie, or chamber western.
Choose the Higgsfield path:

- Image prompt or concept frame: GPT Image 2, Soul Cinema, Soul 2.0, or Nano Banana Pro.
- Serious video or multi-shot scene: Seedance 2.0.
- Fluid action and character movement: Kling 3.0.
- Cinematic hero shot: Cinema Studio / Cinematic Studio when available.
- Product/ad variant: Marketing Studio first, then pulp-cinema prompt polish.
- Write a shot prompt with subject, location, camera, motion, lighting, color, texture, and continuity.
- For longer scenes, split into 10-15 second micro-takes with matching first/last frames.
- Ask before any generation spend unless the user already approved a credit budget.

## Model Router

- Dialogue-heavy scene with two or three people: Seedance 2.0 or Kling 3.0.
- Atmospheric wide shot: Veo / Sora-class model if available, otherwise Seedance 2.0.
- Precise hero still, poster, title card, or visual design: GPT Image 2.
- Character-consistent stills: Soul 2.0 / Soul Cinema with Soul ID if available.
- Fast exploration: Nano Banana Pro or Flux-style image model.
- Finished-video analysis: Virality Predictor.

## Shot Library

### Trunk-Level Reveal

- Use a low camera looking upward from inside a trunk, suitcase, freezer, drawer, or floor-level container.

Prompt pattern:

text

Low-angle camera from inside an open car trunk, two sharply dressed characters leaning into frame, practical overhead light, hard shadows, 35mm vintage prime feel, saturated 1990s crime palette, static camera, tactile props, film grain, no gore.

## Standoff Tableau

Triangular blocking, stillness, eye lines, hands near props, tension before action.

Prompt pattern:

text

Three characters locked in a tense triangular standoff inside a neon-lit warehouse, static wide anamorphic frame, hard rim light, red and amber practicals, hands visible near props, dust in the air, slow push-in, pulp crime tension, no graphic violence.

## Extreme Close-Up Chain

Cut between eyes, hands, boots, prop, clock, mouth, and a final wide reveal.

Prompt pattern:

text

Extreme close-up sequence: eyes narrowing, hand resting beside a diner cup, boot tapping under the table, old wall clock, then a slow reveal of the full booth. Warm tungsten diner light, shallow focus, 1970s film grain.
Walk-And-Talk
Two characters moving in parallel, dialogue implied through posture and rhythm.

Prompt pattern:

text

Two characters walking side by side through a back hallway, camera dolly tracking with them, fluorescent practicals, vintage 35mm lens, casual menace, natural gestures, dialogue-first energy, steady blocking, no random camera cuts.

## Retro Title Card

Short, graphic, chaptered.

Prompt pattern:

text

Black background with mustard-yellow retro serif chapter title text, subtle film gate weave, dust and scratches, 1970s exploitation title-card energy, centered typography, 3-second hold.

## Prompt Formula

Every Higgsfield video prompt should include at least six of these:

- Era or decade
- Location
- Character action
- Camera angle
- Lens or format feel
- Camera motion
- Lighting source
- Color palette
- Texture or film stock feel
- Continuity constraints
- Negative constraints

Template:

text

[Era] [location]. [Characters and action]. Camera: [shot size, angle, lens feel, movement]. Lighting: [source and contrast]. Palette: [colors]. Texture: [grain/halation/softness]. Continuity: [wardrobe, props, positions]. Avoid: warped hands, unreadable text, gore, random cuts, drifting identity.

## Longer Scene Workflow

For 30 seconds to 2 minutes, do not ask one model for the whole film. Build chapters:

- Generate a hero still or storyboard board.
- Split into micro-takes: 5-15 seconds each.
- Define first frame and last frame for every take.
- Reuse wardrobe colors and props as identity anchors.
- Use fewer characters per shot.
- Keep violence implied and brief; use cutaways to props, feet, shadows, doors, or reaction shots.
- After generation, check continuity, hand/face stability, prop drift, camera logic, pacing, and text readability.

## Audio Rules

Describe music by style and source:

- "1960s surf rock guitar from a car radio"
- "1970s funk bass groove with wah guitar"
- "spaghetti-western trumpet and low whistle"
- "warm soul record playing quietly from a jukebox"
- "single-note suspense drone with room tone"
- Do not name copyrighted songs or ask for a specific artist's recording.

## Safety And Style Rules

Do not claim to be Quentin Tarantino.
Do not claim to recreate a living artist's exact style.
Do not generate graphic gore or explicit violence; imply violence through staging, cutaways, and aftermath-free tension.
Do not include real celebrity likenesses unless the user provides rights/permission.
Keep brand/product text legible by using GPT Image 2 for text-heavy frames.

## Verification Checklist

Scene type chosen
Genre module chosen
Higgsfield model selected
Aspect ratio selected
Prompt includes camera, lighting, palette, motion, continuity, and constraints
Characters limited to a manageable count
First/last frames defined for multi-shot scenes
Credit cost estimated before spend
User approval obtained before generation if needed
Output reviewed for continuity and artifacts