Skip to content
VisualCodes
Log inNotify me

Guide · Stable Diffusion

Stable Diffusion prompts

Positive, negative and size, the way your interface expects.

Stable Diffusion is mostly used on your own computer or a service with an interface like Automatic1111, Forge or ComfyUI. It works with two prompts (positive and negative) and size in pixels.

“Try in” → “Stable Diffusion” copies the English prompt with the “Negative prompt:” line and the SDXL size matching the aspect ratio, the format A1111 and Forge read in one go.

Build my promptSee examples

Anatomy of a Stable Diffusion prompt

Each colour is a part the Builder adds for you: no need to remember each tool's format.

Steaming cup of coffee on a table by the window, soft ambient glow, strong background bokeh, cozy intimate moodNegative prompt: textSize: 896x1120

  • Subject and codes
  • Format
  • What must not appear

How to use it with Stable Diffusion

Three steps from the Builder.

  1. Build the combination in the Builder.
  2. Tap “Try in” → “Stable Diffusion”: it's copied with negative and size (no website opens).
  3. In A1111/Forge paste it in the prompt and press the read-parameters button (↙): it splits positive, negative and size.

Ready examples for Stable Diffusion

Top, the VisualCodes combination; bottom, the prompt as the tool receives it. All use free sample codes.

  • A warm, close social post

    Combination

    /AmbientGlow"Steaming cup of coffee on a table by the window"/BokehPortrait/CozyMood-4:5 -no text

    Prompt for Stable Diffusion

    Steaming cup of coffee on a table by the window, soft ambient glow, strong background bokeh, cozy intimate mood Negative prompt: text Size: 896x1120

  • Frozen action against the light

    Combination

    /CityStreet"Skateboarder jumping in a city street"/FreezeAction/Backlight/HighContrastColor-16:9 -no text

    Prompt for Stable Diffusion

    Skateboarder jumping in a city street, city street setting, urban buildings, pavement, traffic cues and everyday street atmosphere, frozen high-speed action, strong backlight, high colour contrast Negative prompt: text Size: 1344x768

  • Drink product with one colour accent

    Combination

    /IceCold"Soda bottle with droplets on a dark background"/RedAccent/Smoke-1:1 -no text -no logo

    Prompt for Stable Diffusion

    Soda bottle with droplets on a dark background, ice-cold drink look, single red accent, soft smoke atmosphere Negative prompt: text, logo Size: 1024x1024

  • Ad story with text

    Combination

    /StoryAd"Sale ad for a clothing store"/ShoppingBag/HighContrastColortext:"-30%" -9:16 -no extra text

    Prompt for Stable Diffusion

    Sale ad for a clothing store, vertical story ad layout, safe interface margins, fast visual hook, clear CTA and offer space, branded retail shopping bag, visible logo placement, premium paper bag mockup, high colour contrast, text: "-30%" Negative prompt: extra text Size: 768x1344

  • Anime-style illustration with an 80s feel

    Combination

    /Anime"Girl with headphones on a street at night"/CityStreet/80s-2:3 -no text

    Prompt for Stable Diffusion

    Girl with headphones on a street at night, anime-inspired style, expressive characters, clean linework, stylised cinematic emotion, polished illustrated look, city street setting, urban buildings, pavement, traffic cues and everyday street atmosphere, 1980s aesthetic, bold colour, neon accents, synth energy, glossy surfaces Negative prompt: text Size: 832x1216

  • Fast food with texture and light

    Combination

    /CrispyEdge"Crispy fries in a paper cone"/Backlight/Joyful-1:1 -no text

    Prompt for Stable Diffusion

    Crispy fries in a paper cone, crispy browned edges, visible crunch texture, appetising cooking contrast, strong backlight, joyful mood, open positive expression, uplifting emotional brightness Negative prompt: text Size: 1024x1024

Where Stable Diffusion works best

Use cases with recommended codes and ready combinations.

Tips

  • In ComfyUI paste the first line in the positive node and the “Negative prompt” line in the negative one.
  • Sizes are the ones recommended for SDXL; with SD 1.5 use roughly half.
  • Add to the negative whatever your model tends to sneak in (odd hands, watermarks).

Common mistakes

  • Writing what you don't want in the positive prompt: it belongs in the negative.
  • Using odd ratios with SDXL: it distorts the composition.
  • Non-English prompts: most models were trained in English.

FAQ

Why doesn't a website open?

Stable Diffusion has no single official website: everyone uses their own interface. That's why the button only copies the prompt.

Does it work with ComfyUI?

Yes: paste each line in its node (positive and negative) and set the size in the empty image node.

What size for a story?

768×1344 (9:16), which the prompt sets if you pick that format in the Builder.

Other tools