Prompt Basics

Why prompts matter

Stable Diffusion is basically “draw what I say” — but it’s picky about how you say it. You don’t need special magic words. You just need to be clear about:

• Who / what is in the image (subject)
• Where they are (scene / background)
• How it should look (style / lighting / mood)

On CloudDock, you type all of that into Prompt, click Generate, and the GPU does the rest in your browser.

These two boxes matter more than any advanced slider. Prompt = what you want. Negative prompt = what you don’t want.

The basic formula

A reliable beginner template looks like this:

[subject / character],
wearing [outfit or look],
in/on [scene / background / situation],
[style keywords / art vibe],
[lighting / color mood / quality words]

Example:

highly detailed anime girl, school uniform, standing on a beach at sunset,
vibrant colors, high contrast, cinematic lighting

That’s plain English. You don’t have to write paragraphs or spam 200 comma words. Stable Diffusion actually does better with clean, readable input than with random buzzword soup.

This output came from the exact prompt above. No secret sauce, no paid model, just the default environment.

Negative prompt (the “please don’t” box)

The Negative prompt field tells the model what to avoid. This is where you fight cursed hands, melted eyes, and random extra limbs.

Copy/paste this as a starter:

blurry, low resolution, extra limbs, extra fingers, deformed hands, distorted face, worst quality

You can keep this almost the same for most images. You tune the main Prompt for style; you keep Negative prompt as a cleanup filter.

Think of Negative prompt as “don’t do this, I mean it.” You can reuse it for almost every shot.

Style & vibe keywords

You can steer the final look just by dropping a few “vibe” words into the Prompt. Here are some that work well in CloudDock’s default setup:

• anime style / anime illustration — softer lines, stylized faces
• vibrant colors / high contrast — saturated, punchy
• cinematic lighting / dramatic rim light — strong light direction, moodier feel
• soft lighting / warm ambient light — gentle, pastel, cozy look
• highly detailed / finely detailed / ultra detailed — adds texture and micro detail

Don’t stack every word. Pick 2–4 that match what you want. Too many descriptors can backfire and make the image noisy.

“vibrant colors, high contrast, cinematic lighting” → punchy and dramatic.

“soft lighting, warm ambient light, pastel colors” → gentle and warm.

Pose, framing, camera angle

You can hint at pose and framing by literally saying camera-ish stuff:

full body shot, standing pose, facing forward, hands visible

waist-up portrait, looking at viewer, slight smile

dynamic angle from below, windswept hair, action pose

These don’t guarantee perfection (hands are still… hands), but they do nudge the model in the right direction. For beginners, “full body shot” vs “waist-up portrait” alone already makes a big difference.

Left: “waist-up portrait, looking at viewer.” Right: “full body shot, standing pose.” Same character, different framing.

When it’s not listening

It drew the wrong outfit / wrong background

Put the important part earlier in the prompt. The model “pays more attention” to the front.

school uniform, highly detailed anime girl, standing on a beach at sunset ...

vs

highly detailed anime girl, standing on a beach at sunset, wearing a school uniform ...

The first version makes “school uniform” higher priority.

The hands / fingers are still cursed

Totally normal. Hands are where everyone suffers. Add hands visible or hands in frame to the Prompt so the model understands you care, and keep extra fingers, deformed hands in your Negative prompt.

Later we’ll show you how to clean this up even more in the “Fix blurry faces / bad hands” guide.

The face is mushy / low detail

Try “highly detailed face, sharp eyes” in the Prompt. Do not immediately crank resolution to 4K as step one — that just murders VRAM.

Copy/Paste starter recipes

Beach anime poster style

Prompt:

highly detailed anime girl, school uniform, standing on a beach at sunset,
vibrant colors, high contrast, cinematic lighting, full body shot, looking at viewer

Negative prompt:

blurry, low resolution, extra limbs, extra fingers, deformed hands, distorted face, worst quality

High-contrast beach poster style. This is the “show-off to your friends” one.

Soft pastel / comfy vibe

Prompt:

cute anime girl, oversized hoodie, sitting indoors by a window,
soft lighting, pastel colors, warm ambient light, waist-up portrait, gentle expression

Negative prompt:

blurry, harsh shadows, overexposed, extra limbs, extra fingers, worst quality

Pastel “cozy streamer thumbnail” vibe. Warm light, soft edges, less drama.

What’s next?

At this point you can already steer character, pose, and vibe using plain English. From here you’ve got two big upgrade paths:

• Fix blurry faces / cursed hands: Learn a couple tiny slider tweaks and resolution tips to make things cleaner without nuking VRAM.
  Go to “Fixing Blurry Faces & Bad Hands →”
• Use a custom style / character model: Either pick a pre-trained style from the Model Library, or train your own character with our Training Center so it remembers your brand / VTuber look.
  Go to “Model Library Overview →”