🧠 Perfect Prompting: The GPT Engine for SDXL, Flux & Pony
Created by Rhix
https://chatgpt.com/g/g-687ab836709081919c7e98d935b50335-perfect-prompting
A precision‑engineered assistant for crafting high‑quality, layered prompts for Stable Diffusion XL (SDXL) | PONY | FLUX — ideal for Civitai, ComfyUI, A1111, and InvokeAI. Whether you're pursuing cinematic realism, stylized anime, fantasy, abstract art, or creative experimentation — this is your go‑to tool.
🔧 Core Capabilities
Prompt Layering
– Modular, editable structure using clearly labeled layers:
[Main Subject] → [Pose] → [Lighting & Background] → [Technical Style Descriptors] → [Quality Boosts] → [Reference Styles]
Allows easy refinement, reordering, and weights insertion (e.g. (word:1.4)).
Structured Prompt Output
– One‑line positive prompts tailored for SDXL
– Separate high‑quality negative prompts to exclude flaws
– Clean SCSS‑style code blocks, ready to paste into your UI
Default Negative Prompt Baseline
text, logo, watermark, patreon username, deformed hands, extra fingers, unnatural eyes, asymmetric face, blurry, oversaturated, low contrast, jpeg artifacts
Cinematic & Anime Prioritization
– Suitable for photorealistic cinematography or classic anime styles
– Emphasis on accurate anatomy, expressive emotion, depth, and composition
Photography-Inspired Vocabulary
Use terms like:
rim lighting, rule of thirds, soft shadows, wide-angle, shallow depth of field, volumetric light, bokeh, 85mm lens effect—to enhance realism and visual storytelling
🌐 Community-Aware & Trends-Sensitive
– Monitors Civitai keyword trends, high‑performing tags, and styles
– Integrates popular LoRAs, aesthetics, and structured prompts from the community
– Supports prompt weighting, contrast layering, and seed control for reproducibility
🛠 Smart Collaboration Features
Feature Benefit Iterative brainstorming Provide a simple idea and get an optimized full prompt Prompt refinement Upload or type your old prompt—get a cleaner, better‑layered version Style reverse‑engineering Use image analysis to derive matching prompt structure and style tags
🆕 Changelog (Version 1 → Current)
1.1: Improved negative baseline; added wat rtnaming filter
1.2: Enhanced photography terms — added bokeh, Rim and volumetric lighting
1.3: Weighting support
(word:1.4)throughout layering1.4: Deepened Civitai trend integration — auto‑suggest styles
1.5: Prompt reverse‑engineering from image input
1.6 (Current):
Better anime support: clean lines, expressive eyes
Expanded embedding support:
<lora:...>and community embedding tagsRefined post‑processing: auto‑detect key tags for Civitai upload
https://chatgpt.com/g/g-687ab836709081919c7e98d935b50335-perfect-prompting
💡 How to Use Perfect Prompting
1. Ideation
Describe your concept: e.g., “mysterious cyberpunk samurai in neon rain” or “realistic portrait of an elf warrior”.
2. Prompt Layering (Example)
[Main Subject]: cyberpunk samurai, futuristic armour
[Pose & Perspective]: 3/4 view, sword raised
[Lighting & Background]: neon rain‑lit street, reflections on wet pavement, rim light
[Style & Technical Descriptors]: ultra detailed, sharp focus, realistic skin, shallow depth of field
[Quality & Boost Keywords]: masterpiece, best quality, 8k, RAW photo
[Reference Styles]: near Ridley Scott aesthetic, cinematic lighting, Civitai trending tag “neon samurai”
3. Final Structured Output
"cyberpunk samurai in futuristic armour, 3/4 perspective, sword raised, neon rain‑lit street with wet pavement reflections, rim lighting, ultra detailed realistic skin, shallow depth of field, masterpiece, best quality, 8k, RAW photo, cinematic lighting"
Negative Prompt
text, logo, watermark, deformed hands, extra fingers, unnatural eyes, messy anatomy, low quality
4. Refine & Experiment
– Adjust focus by moving modifiers or adding (word:1.3) weights
– Swap layers (e.g., try side view instead of 3/4)
– Convert prompts to Flux Realistic or Pony Diffusion XL formats effortlessly
🧠 Model-Specific Support
SDXL
– Use photography vocabulary, one-line prompt, full-weighting, community tags
– Best used with ComfyUI, A1111, InvokeAI
Flux Realistic / Flux Photoreal
Positive Prompt: masterpiece, best quality, photo‑realistic, sharp focus, ultra‑detailed, dramatic lighting, bokeh, soft shadows, natural light, 85mm lens
Pony Diffusion XL
– Use full‑quality modifiers like score_9, score_8_up, etc.
– Start with source_pony, anthro pony or source_cartoon pony, then rest of prompt
✅ Posting to Civitai
When you're happy with a prompt, Perfect Prompting can auto-generate metadata for your Civitai submission:
Suggested Title: e.g. “Neon Samurai in Rain – 8k Cinematic SDXL Prompt”
Description: concise breakdown of layers, style references, lighting, and how to tweak
Tags: trending Civitai tags, community LoRAs, style keywords
Supports preview images and credit formatting
✨ Why Choose Perfect Prompting?
Modular, editable prompt structure — easily tweak and evolve your ideas
High‑quality realism with attention to lighting, anatomy, cinematic composition
Community‑driven: aligned with Civitai trends and popular embeddings
Cross‑model flexibility: works with SDXL, Flux, Pony Diffusion and beyond
Iterative collaboration: from raw idea → polished prompt → Civitai‑ready submission
🚀 Quick Start Example
You say: “I want a vibrant anime scene: flying futuristic dragon over neon Tokyo by night.”
Perfect Prompting breaks it down:
[Main Subject]: flying futuristic dragon [Pose]: aerial view, wings spread [Lighting & Background]: neon Tokyo skyline at night, glowing signs, volumetric light [Style]: anime style, clean lines, expressive face [Quality]: masterpiece, ultra detailed, hyperrealistic, cinematic lighting [Reference]: trending “cyber dragon Tokyo” tag on CivitaiGet final positive and negative prompts, try generation, adjust weights or lens focus, export title/tags for upload.
Ready to generate your first masterpiece? Just share your idea, and we’ll build the perfect SDXL prompt together. 🌟
https://chatgpt.com/g/g-687ab836709081919c7e98d935b50335-perfect-prompting