Stable Diffusion Prompt Generator: SDXL, Pony V6 & Hand Fixes
Stop generating "burned" images and 7-fingered hands. The Promptsera Prompt Engineer is the only tool updated for Pony Diffusion V6 (Score Syntax), SDXL Turbo, and Stable Diffusion 3. We automatically format your ideas with the correct weights, embeddings, and resolution settings.
Auto-injects
score_9, score_8_up tags required for high-quality anime generation.One-click access to "EasyNegative" and "BadHandV4" strings to fix anatomy instantly.
Applies A1111/ComfyUI syntax
(keyword:1.2) to emphasize key elements properly.Generate Your Optimized Prompt
Select your Model Architecture: SDXL (Realism) or Pony (Anime/Stylized).
Stable Diffusion Prompt Generator
Create positive and negative prompts tailored for diffusion models.
Stable Diffusion Prompts
Compatible with: Automatic1111, ComfyUI, Forge, Fooocus, and Civitai Generator.
Mastering Pony Diffusion V6: The "Score" System
If you are switching from SD 1.5 to Pony Diffusion V6 (or NAI-based models), your old prompts will fail. Pony ignores standard quality tags like "masterpiece." Instead, it was trained on a strict Score Ranking System.
To get usable images, you must prefix your prompt with the following tags, which our generator adds automatically in "Anime" mode:
- score_9: Targets the top 10% of aesthetic quality in the dataset.
- score_8_up: Includes high-quality images (crucial for detail).
- source_anime / source_pony: Tells the model which art style to emulate.
SDXL vs. SD 1.5: Why Your Prompts Look Different
Stable Diffusion XL (SDXL) uses a dual text-encoder (OpenCLIP + CLIP ViT-L). This means it understands Natural Language much better than the "Tag Soup" style of SD 1.5.
| Feature | Stable Diffusion 1.5 | SDXL / SDXL Turbo |
|---|---|---|
| Prompt Style | Comma-separated tags (Danbooru) | Full descriptive sentences |
| Resolution | 512x512 (Native) | 1024x1024 (Native) |
| Negative Prompts | Heavy reliance (EasyNegative, etc.) | Minimal reliance (Too many breaks the image) |
Prompt Weighting Syntax Explained: ( ) vs [ ]
One of the most common questions we get is: "What do the parentheses mean?" In Automatic1111 and ComfyUI, this is how you control the AI's attention.
(keyword): Increases attention by a factor of 1.1x.((keyword)): Increases attention by 1.21x (Nested).(keyword:1.5): Manually sets attention to 150%. Use this for the main subject.[keyword]: Decreases attention by 0.9x. Use this for elements you want to fade into the background.
Pro Tip: Never set a weight above 1.6 or below 0.4, as this will usually "burn" the image or create artifacts.
📚 Universal Negative Prompt Library (Copy & Paste)
The secret to fixing "bad hands" and "extra limbs" isn't just luck; it's using the right negative embeddings.
Option 1: The "Universal Fixer" (Best for SD 1.5)
(worst quality, low quality:1.4), bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, jpeg artifacts, signature, watermark, username, blurry, artist name, bad legs, error, malformed limbs, fused fingers, too many fingers, long neck, cross-eyed, mutated hands, polar lowres, bad body, bad proportions, gross proportions.Option 2: The "SDXL Safe" (Best for Juggernaut/Realism)
(worst quality:1.2), (low quality:1.2), (lowres:1.1), (monochrome:1.1), (greyscale), multiple views, comic, sketch, (blurry:1.15), (text:1.1), watermark, logo.Option 3: The "Pony Specific" (Preventing Stylized Trash)
score_4, score_5, score_6, source_furry, source_pony, 3d, cgi, render, bad anatomy, bad hands, text, watermark, simple background.Troubleshooting: Why Do I Have "Double Heads"?
The most common error in Stable Diffusion is the "Two-Headed Glitch." This happens when you generate an image larger than the model's native training resolution. The AI tries to fill the extra space by simply repeating the subject.
Resolution Cheat Sheet
- Stable Diffusion 1.5: Stay near
512x512or512x768. (Use "Hires Fix" to go bigger). - SDXL 1.0 / Pony: Stay near
1024x1024or832x1216. Never generate 512x512 with SDXL; it will look blurry and broken.
Frequently Asked Questions
How do I fix faces in Stable Diffusion?
For wide shots where faces look distorted, use the ADetailer extension in Automatic1111. It automatically detects the face after generation, masks it, and regenerates just the face at a higher resolution. Prompting alone often cannot fix small faces.
What does (keyword:1.2) mean?
This syntax tells the AI to pay 20% more attention to that specific keyword. It is useful for forcing specific colors or objects that the AI is ignoring.
Why do my Pony Diffusion V6 images look low quality?
You likely forgot the "Score Tags." Pony Diffusion V6 requires score_9, score_8_up, score_7_up at the very start of the positive prompt. Without these, the model assumes you want low-quality output.
Expand Your AI Toolset
Promptsera is the leading suite for free AI utilities. Once you master Stable Diffusion, try our other specialized tools.