Stable Diffusion Prompt Optimizer

Stop fighting with syntax errors in WebUI. Visually build dense, heavily weighted Positive and Negative prompts optimized for SDXL and SD 1.5 models.

7

How closely the AI follows the prompt (7.0 is standard).

20

More steps = sharper image but slower generation.

Positive Prompt

A lone futuristic samurai walking down an alleyway, (masterpiece:1.2), (best quality:1.1), (ultra highres:1.2), 8k resolution, highly detailed, HDR, cinematic lighting, volumetric lighting, ray tracing, glowing ambient light

Negative Prompt

nsfw, lowest quality, worst quality, low resolution, blurry, mutated, deformed, text, watermark

Steps: 20CFG: 7Sampler: Euler a

The Secret to Stable Diffusion

If you are moving from DALL-E or Midjourney into the open-source world of Stable Diffusion (using local UIs like Automatic1111 or ComfyUI), you quickly realize it doesn't understand natural English sentences well. Instead, to get photographic perfection, you must learn the art of "Keyword Weighting."

How Weighting Works

  • The Standard: In Stable Diffusion, keywords placed at the beginning of the prompt hold more mathematical strength than words at the end.
  • The Parentheses (): Wrapping a word in parentheses (e.g., (masterpiece)) multiplies its strength by 1.1x.
  • The Colon :: Adding a colon and a number allows fine-tuned granular control. To make a subject have extreme lighting, you would write (cinematic lighting:1.3).

Negative Prompts Are Mandatory

Midjourney automatically applies thousands of "hidden" rules to prevent its model from generating ugly images. Because Stable Diffusion is raw and open-source, it will readily generate horrifying logic (like subjects with 7 fingers) unless you explicitly tell it not to via a massive Negative Prompt block. This tool sets up an optimal default negative block for you instantly.