Midjourney v7 Cinematic Prompt: abandoned Victorian greenhouse overtaken by vines in claymation stop-motion still

You are generating a single production-grade image prompt for Midjourney v7. Output the prompt only — no preamble.

## Scene
- **Subject:** abandoned Victorian greenhouse overtaken by vines
- **Style:** claymation stop-motion still
- **Lighting:** overcast diffuse soft
- **Composition:** golden ratio spiral
- **Aspect ratio:** --ar 3:2

## Midjourney v7 Prompt Architecture

### Structure (in order)
1. **Medium** (photograph / oil painting / 3D render / illustration)
2. **Subject** + action pose
3. **Environment** + 2-3 concrete details
4. **Lighting** (overcast diffuse soft)
5. **Style anchors** (claymation stop-motion still)
6. **Technical quality tags**
7. **Model-specific parameters**

### Model-specific formatting

**Midjourney v6.1 / v7:**
- Natural language, comma-separated clauses
- Camera body + lens: "shot on Hasselblad H6D, 80mm lens, f/1.8"
- Parameters after a double-dash: `--ar 3:2 --stylize 250`
- v6+ handles quoted text well: `"NEON" sign glowing`
- Avoid weight-syntax `::` unless splitting concepts

**Flux.1 [pro] / [dev] / [schnell]:**
- Flux is the reigning champion of text rendering and anatomy
- Use natural long-form sentences, not tag soup
- Does NOT use Midjourney-style parameters
- Prompt length sweet spot: 60-120 words
- "[pro]" for final quality, "[dev]" for variations, "[schnell]" for speed

**SDXL / SD 3.5 Large:**
- Tag-style prompts with weighting: `(cinematic lighting:1.3), (film grain:0.8)`
- Trigger words matter if using LoRAs
- Split positive / negative prompts
- CFG 5-8 for SDXL; 4-6 for SD3.5

**Ideogram 2.0:**
- KING of in-image text. Lean in: "a poster reading 'CLOSING SOON' in bold serif".
- Natural language; supports magic prompt
- Style presets: Design / Realistic / Anime

**Recraft V3:**
- Vector-first; great for logos, icons, illustrations
- Use "vector illustration" or "flat design" explicitly
- Brand palette support via their UI

**DALL-E 3 / ChatGPT Image:**
- Rewrites your prompt internally — keep intent clear, don't over-specify pixel details
- Excellent at following compositional instructions
- Aspect ratio via natural language: "wide landscape format"

**Imagen 3:**
- Photorealism leader alongside Flux
- Responds to camera/lens specs like Midjourney
- Available via Gemini API

**Nano Banana (Gemini 2.5 Flash Image):**
- Insanely fast; great for iteration
- Supports image-in-image editing natively
- Use conversational prompts: "change the sky to golden hour, keep everything else identical"

**niji v6:**
- Anime/manga only
- Use anime-specific tags: "cel shading", "shoujo style", "bishoujo", "1girl/1boy"

**Playground v3:**
- Best-in-class for editorial and fashion imagery
- Strong on faces and skin texture

## Output

**A. Primary Prompt** (in code block, 60-120 words for Flux/Imagen/DALL-E; 40-80 for Midjourney/SDXL):
```
[Your prompt here, formatted for Midjourney v7]
```

**B. Parameters line** (model-specific):
For Midjourney: `--ar 3:2 --stylize 250`
For SDXL: CFG, sampler, steps, scheduler
For Flux: "[dev]" or "[pro]", guidance scale 3.5
For DALL-E 3: "natural" or "vivid" style

**C. Negative prompt** (if Midjourney v7 uses one — SDXL yes, Midjourney via `--no`, Flux no):
`[standard negatives for claymation stop-motion still]`

**D. Variation hints** (3 bullets):
- 2 alternate overcast diffuse soft options
- 1 alternate golden ratio spiral
- 1 parameter tweak to push style stronger

## Pitfalls
- Do not use comma-chains longer than 15 items on Flux / Imagen — they get confused.
- Midjourney ignores `photorealistic` — use `--style raw` instead.
- SDXL absolutely needs a strong negative prompt; Flux does not use negatives.
- Ideogram's text rendering fails above ~12 characters reliably — keep signage short.
- Midjourney v7 specifically: [insert one quirk relevant to Midjourney v7]

Produce the output.