Phase 6: Image, Video & Voice Prompting · 30 min · Midjourney, DALL-E 3, or Stable Diffusion

Negative Prompts & Style Consistency

Telling the AI what to avoid is half the battle. Keeping style consistent across images is the other half.

Hiring signal: What changes when you use this: you can produce a consistent set of images that look like they belong together — the skill that turns AI images from one-offs into campaign-ready assets.

What you will learn

Use negative prompts effectively — short, specific, 3-5 terms max
Iterate on a visual style and keep it consistent across a set of images
Build a style guide prompt that can be reused across multiple images
Understand which tools support negative prompts and how to use them

Negative Prompts

Negative prompts tell the AI what you DON'T want in the image. They're available in Midjourney (using --no), Stable Diffusion (negative prompt field), and some other tools.

When to use negative prompts:

The AI keeps adding something you don't want (text, watermarks, extra fingers)
The style keeps drifting in a direction you don't want
You want to exclude specific elements

How to write good negative prompts:

Short — 3-5 terms max
Specific — "text, watermark, signature" not "bad stuff"
Concrete — "blurry, low quality, distorted faces" not "ugly"

Example:

Positive: A serene mountain lake at sunrise, crystal clear water reflecting snow-capped peaks, golden light, cinematic photography, shot on 35mm film
Negative (or --no in Midjourney): text, watermark, people, buildings, boats, harsh shadows

Without the negative prompt, the AI might add a hiker on the shore or a boat on the lake — because that's what it's seen in mountain lake photos. The negative prompt keeps it pristine.

Negative prompts are for recurring problems, not wishlists
Don't use negative prompts to list everything you might not want. Use them for things the AI keeps adding that you don't want. If every image has unwanted text, add "text, watermark" to the negative prompt. If the faces look distorted, add "distorted, deformed." Keep it focused on actual recurring issues.

Style Consistency

When you need a set of images that look like they belong together (a campaign, a series, a brand), you need style consistency. Here's how:

1. Build a Style Anchor

Write a "style suffix" — a block of text that describes the visual style, appended to every prompt:

Style anchor: "Cinematic photography, shot on 35mm film, Kodak Portra 400, warm color grade, soft natural lighting, shallow depth of field, slight film grain, minimal composition with negative space"

2. Use the Same Style Anchor for Every Image

Image 1: A woman reading in a park. [style anchor] Image 2: A man walking a dog on a city street. [style anchor] Image 3: A child playing in a garden. [style anchor]

All three images will share the same visual style because the style anchor is identical.

3. In Midjourney: Use `--sref` (Style Reference)

Midjourney lets you reference a style from a previous image:

[subject description] --sref [image URL or code]

This keeps the style consistent across generations.

4. Document Your Style Guide

Write down your style anchor, your negative prompt, and any tool-specific settings (aspect ratio, quality, etc.). This is your style guide — anyone should be able to use it and get consistent results.

Unlock the full lesson

You've read the first 2 sections. The rest of this lesson covers Your Practice Rep — plus a hands-on lab, quiz, and project artifact.

Create a free account to unlock Phase 0 and Phase 1 of every course — no credit card.