superimage-generator

Files (72)

• SKILL.md
• INDEX.md · examples/
• asian-bar-candid-dreamy.json · examples/
• backyard-lounger.json · examples/
• balcony-flash-selfie.json · examples/
• beach-cabana-turquoise-bikini.json · examples/
• bedroom-floral-sleepwear.json · examples/
• black-athleisure-thumbs-up.json · examples/
• brand-concept-object-editorial-template.json · examples/
• brand-social-collage-campaign-template.json · examples/
• bw-window-portrait.json · examples/
• cafe-sideways-selfie-thigh.json · examples/
• car-rainy-night.json · examples/
• champagne-lace-elevator-selfie.json · examples/
• coquette-bedroom.json · examples/
• editorial-portrait.json · examples/
• elevator-mirror-selfie.json · examples/
• fashion-editorial-split-catalog.json · examples/
• fashion-studio-collage.json · examples/
• gym-bench-vtaper.json · examples/
• hawaiian-tropic-loungewear.json · examples/
• heart-pendant-hallway-selfie.json · examples/
• helicopter-coastline-portrait.json · examples/
• hotel-bathroom-couple.json · examples/
• iris-necklace-portrait.json · examples/
• ivory-nightdress-flash-pov.json · examples/
• jet-ski-lookback.json · examples/
• kitchen-candid.json · examples/
• lightning-mcqueen-bathroom-selfie.json · examples/
• low-angle-lace-bedroom-portrait.json · examples/
• macbook-screen-nested.json · examples/
• mermaid-gown-editorial.json · examples/
• mirror-halter-denim-side-profile.json · examples/
• monaco-rooftop.json · examples/
• morning-bed-selfie-oversized-tee.json · examples/
• navi-beach-night-flash.json · examples/
• outdoor-cafe-athletic-collage.json · examples/
• pool-night-lounging.json · examples/
• poolside-lip-gloss.json · examples/
• product-serum-mirror-slab.json · examples/
• red-hair-corset-editorial.json · examples/
• reference-image-gym.json · examples/
• seaside-terrace.json · examples/
• snow-forest-playful.json · examples/
• spongebob-shorts-bathroom-mirror.json · examples/
• streetwear-editorial-collage.json · examples/
• striped-shirt-ponytail-mirror-selfie.json · examples/
• tropical-beach-selfie.json · examples/
• venetian-blinds-athletic.json · examples/
• vintage-corvette.json · examples/
• white-backless-dress-hallway.json · examples/
• white-bikini-mediterranean-terrace.json · examples/
• wink-tongue-out-black-dress-studio.json · examples/
• workout-shorts-mirror.json · examples/
• xbox-green-bikini-bedroom-selfie.json · examples/
• black-forest-labs.md · models/
• flux-kontext-max.md · models/
• flux.md · models/
• gpt-image-1.5.md · models/
• grok-imagine.md · models/
• …and 12 more

SKILL.md (first lines)

# Superprompt Generator

Transform basic prompts into professional-grade superprompts optimized for specific AI image models.

**Terminology:**

- **prompt** = user's basic input (e.g., "a woman taking a selfie")
- **superprompt** = the enhanced, structured output with detailed specifications

## ⚠️ MANDATORY WORKFLOW - DO NOT SKIP ANY STEPS

This skill has a **sequential, dependency-based workflow**. Each step builds on the previous one. **Skipping steps produces inferior results and failed generations.**

### Critical Enforcement Rules:

- [ ] **Step 1 MUST complete before Step 2** — No model selection without intent clarification
- [ ] **Step 3 MUST check `examples/INDEX.md` before generating** — Read the index, follow its instructions. Do NOT read more than 2 example files.
- [ ] **Step 4 VALIDATION is mandatory — no exceptions** — Run validation script before proceeding
- [ ] **Do NOT flatten (Step 5) until validation returns `✓ VALID`** — Invalid prompts fail on models
- [ ] **Always show user the validation report** — They need to see quality scores and any warnings

**If you are tempted to skip steps: DON'T. The workflow exists for a reason.**

## Contents

- [Quick Start](#quick-start)
- [Workflow Selection](#workflow-selection)
- [Workflow](#workflow)
- [Reference Image Workflow](#reference-image-workflow)
- [Superprompt Playbook](#superprompt-playbook) (see also [playbook.md](playbook.md))
- [Examples](#examples)
- [Model Templates](#model-templates)
- [Validation Scripts](#validation-scripts)
- [Aspect Ratio Reference](#aspect-ratio-reference)

## Quick Start

> [! IMPORTANT]
> Do not act after reading the quick start section alone. Make sure to read [Workflow Selection](#workflow-selection) before commiting to generating an output.

**Input:** `"a woman taking a selfie"`

**Output:** Structured superprompt with `subject`, `wardrobe`, `pose`, `scene`, `lighting`, `camera` blocks containing anatomical precision, fabric details, camera technicals, and negative constraints.

**With Reference Image:** `"gym selfie" + [uploaded face reference]`
→ See [Reference Image Workflow](#reference-image-workflow) below.

## Workflow Selection

**Does the user provide a reference image for face/identity preservation?**

- **YES** → Follow [Reference Image Workflow](#reference-image-workflow)
- **NO** → Follow [Standard Workflow](#workflow) below

---

## Workflow

Copy this checklist and track progress — **each step is mandatory**::

```
Superprompt Generation (SEQUENTIAL):
- [ ] Step 1: Clarify user intent ← START HERE
- [ ] Step 2: Select target model ← DEPENDS ON STEP 1
- [ ] Step 3: Generate structured superprompt ← DEPENDS ON STEP 2
- [ ] Step 4: Validate with script ← MANDATORY, NON-OPTIONAL
- [ ] Step 5: Flatten for model format ← ONLY AFTER VALIDATION PASSES
- [ ] Step 6: Return to user ← FINAL STEP
```

### Step 1: Clarify Intent

**DO NOT skip this step. If unsure about intent, ask the user:**

- Main subject? (person, product, landscape, architecture, food, animal)
- Style? (photorealistic, artistic, illustration)
- Aspect ratio preference? (9:16 mobile, 16:9 landscape, 1:1 square)
- Any other specific mood or setting?

**Document their answers. You need these for Step 2.**

### Step 2: Select Model

**Model selection depends on Step 1 answers. Use the table below:**

**Defaults by subject type:**

- **Person/human subject** → `nano-banana` (anatomical precision, complex poses)
- **Product/object/landscape** → `flux` (general photorealism)

When to use a different model:

- Need text rendering? → `ideogram` or `gpt-image-1.5`
- Need face/identity preservation? → `flux-kontext-max`
- Need artistic/aesthetic style? → `midjourney`
- Need video/motion? → `luma`
- Need iterative editing? → `gpt-image-1.5`

| Model              | Best For                                           | Template                                                 |
| ------------------ | -------------------------------------------------- | -------------------------------------------------------- |
| `nano-banana`      | **Default (people)** - Anatomical precision, poses | [models/nano-banana.md](models/nano-banana.md)           |
| `flux`             | **Default (non-people)** - General photorealism    | [models/flux.md](models/flux.md)                         |
| `flux-kontext-max` | Image editing, identity preservation               | [models/flux-kontext-max.md](models/flux-kontext-max.md) |
| `gpt-image-1.5`    | Reasoning-based, UI mockups, logos                 | [models/gpt-image-1.5.md](models/gpt-image-1.5.md)       |
| `seedream-4.5`     | Natural language, text in images, 4K               | [models/seedream-4.5.md](models/seedream-4.5.md)         |
| `midjourney`       | Aesthetic quality, artistic                        | [models/midjourney.md](models/midjourney.md)             |
| `ideogram`         | Typography, logos                                  | [models/ideogram.md](models/ideogram.md)                 |
| `recraft`          | Illustrations, vectors                             | [models/recraft.md](models/recraft.md)                   |
| `luma`             | Video, 3D, motion                                  | [models/luma.md](models/luma.md)                         |
| `reve`             | Artistic styles                                    | [models/reve.md](models/reve.md)                         |
| `grok-imagine`     | Quick iterations                                   | [models/grok-imagine.md](models/grok-imagine.md)         |

**If uncertain which model to use:** Ask the user about subject type (person vs. object/scene) and desired style, then apply the defaults above.

### Step 3: Generate Superprompt

**Before generating:**

1. Open `examples/INDEX.md`

From BluebagAI/skills, Apache-2.0.

More from BluebagAI/skills