Mastering Negative Prompts and Aspect Ratios for Precise AI Image Generation

This article teaches how to use uppercase NO, without, avoid, or exclude keywords to prevent unwanted elements in AI‑generated images and how to select the proper size parameter to match the desired aspect ratio, with concrete examples and best‑practice formulas.

James' Growth Diary
James' Growth Diary
James' Growth Diary
Mastering Negative Prompts and Aspect Ratios for Precise AI Image Generation

Hello, I'm James.

In the previous article we listed 10 common mistakes; today we cover two advanced techniques for precise control: negative constraints and aspect‑ratio control.

“I generated a satisfying landscape, but a random person appeared in the lower‑right corner.”
“I generated a cover image, but the ratio is square, leaving white borders on the sides of the banner.”

01 | Why AI adds unwanted elements

GPT Image 2 is trained on a massive image corpus and fills scenes with what it deems reasonable. For example, prompting "a coffee shop" will automatically include tables, lights, windows, and even faint customers because most coffee‑shop pictures contain them. This is a feature, not a bug: the model fills what it thinks is reasonable, which may not match the user's intent.

Typical problems: wanting a minimalist style but getting many decorations; wanting an empty scene but getting crowds; wanting a flat illustration but getting shadows and depth.

Solution: explicitly tell the model what NOT to include.

02 | Negative constraint syntax: explicit exclusion

Use natural language at the end of the prompt with uppercase NO or words like without, avoid, exclude.

# ❌ No exclusion, result uncontrolled
a cozy coffee shop interior, warm lighting

# ✅ With exclusion
a cozy coffee shop interior, warm lighting
NO people, NO customers, NO text signs, NO cluttered background

Common forms:

# Method 1: NO + element (most concise)
NO people, NO shadows, NO text

# Method 2: without
a clean workspace without any clutter, without monitors, without paper

# Method 3: avoid
avoid realistic textures, avoid dark background, avoid complex details

# Method 4: do not include (most explicit)
do not include any human figures, do not add watermarks

Before/after test:

# Before (no exclusion)
prompt: "a minimalist desk setup"
Result: desk with monitor, keyboard, cup, plant…

# After (with exclusion)
prompt: "a minimalist desk setup, clean surface
NO monitor, NO keyboard, NO accessories, only a single notebook"
Result: only a notebook on a clean desk
Tip: place exclusion words at the end of the prompt for better effect.

03 | Advanced negative constraints: three‑layer exclusion

When finer control is needed, write exclusions in three dimensions.

First layer – element

NO watermark, NO text, NO logo, NO people, NO animals
NO cars, NO buildings, NO clutter, NO shadows

Second layer – style

For flat illustration, exclude photorealistic traits:

# Flat illustration style
flat illustration style
NO photorealistic textures, NO 3D rendering, NO shadows, NO gradients
NO depth of field blur, NO lens flare

# Minimal line art
minimal line art style
NO color fills, NO shading, NO background details, NO noise

Third layer – scene elements

# Indoor scene, avoid outdoor view
interior scene of a library
NO windows showing outdoor view, NO natural light source visible
NO outdoor trees or sky visible through windows

# Product photo, avoid background clutter
product photography of a blue sneaker
NO human feet, NO other products nearby, NO background objects
pure white background only

Combined example:

a flat vector illustration of a city skyline at dusk
warm orange and purple tones
NO people, NO cars, NO detailed windows
NO photorealistic rendering, NO shadows, NO gradients
NO text or labels, NO watermarks
clean geometric shapes only

04 | Aspect‑ratio basics: the size parameter in GPT Image 2

GPT Image 2 supports three fixed size values: 1024x1024 – 1:1 square – suitable for avatars, stickers, social‑media posts. 1536x1024 – 3:2 landscape – suitable for article illustrations, banner headers, desktop wallpapers. 1024x1536 – 2:3 portrait – suitable for phone wallpapers, vertical posters, public‑account covers.

⚠️ GPT Image 2 does not support arbitrary ratios such as 16:9; you can describe the desired composition in the prompt to approximate the visual effect.

API example:

# Landscape image (article illustration)
response = client.images.generate(
    model="gpt-image-alpha",
    prompt="your prompt here",
    size="1536x1024",
    n=1
)

# Portrait image (public‑account cover)
response = client.images.generate(
    model="gpt-image-alpha",
    prompt="your prompt here",
    size="1024x1536",
    n=1
)

# Square image (avatar/sticker)
response = client.images.generate(
    model="gpt-image-alpha",
    prompt="your prompt here",
    size="1024x1024",
    n=1
)

05 | Matching content to aspect ratio

Aspect ratio influences composition, visual focus, and information density.

1:1 square → social media, avatars, stickers

Suitable for:
- Instagram post image
- WeChat avatar
- Emoji/sticker
- Product showcase with white margins

Prompt tips:
- Center the subject, leave margin
- Use "centered composition, square format"

Example:
"a cute cartoon cat sitting in the center
centered composition, pastel background
NO text, NO border, square format"

1536x1024 landscape → article illustration, banner

Suitable for:
- Public‑account article image
- Website banner
- Desktop wallpaper
- Presentation background

Prompt tips:
- Use "wide landscape composition" or "horizontal layout"
- Allow side margins for text

Example:
"a wide aerial view of a misty mountain forest at dawn
panoramic landscape composition, horizontal format
golden hour lighting, NO people, NO buildings"

1024x1536 portrait → phone wallpaper, vertical poster

Suitable for:
- Public‑account cover (recommended size)
- Phone lock screen
- Vertical poster/advert
- Book cover

Prompt tips:
- Use "vertical composition, portrait orientation"
- Reserve upper or lower area for title/text

Example:
"a vertical poster design featuring a lone tree on a hill
portrait orientation, vertical composition
dramatic sunset sky filling the upper half
NO text, NO watermarks, clean artistic style"

Quick reference list:

Portraits – portrait – 1024x1536 Landscapes – landscape – 1536x1024 Product shots – square – 1024x1024 Infographics – portrait – 1024x1536 Scene illustrations – landscape – 1536x1024 Avatars/stickers – square –

1024x1024

06 | Combining aspect ratio and negative constraints

Using both techniques yields precise control.

Case 1: Public‑account article cover (portrait)

# Goal: clean tech‑themed vertical poster, no clutter
"a clean tech-themed vertical poster
deep navy blue background with subtle circuit board pattern
centered glowing white geometric symbol in the middle
portrait orientation, vertical composition
NO text, NO people, NO photorealistic rendering
NO cluttered details, NO lens flare, NO watermarks
minimal and elegant style"
size: "1024x1536"

Case 2: Article illustration showing before/after (landscape)

# Goal: contrast "with exclusion" vs "without exclusion"
"a split illustration showing two coffee shop scenes side by side
left side: busy crowded coffee shop with many people and objects
right side: minimalist empty coffee shop, clean and simple
horizontal layout, wide format
flat illustration style, warm colors
clear visual contrast between cluttered and minimal"
size: "1536x1024"

Case 3: Product showcase (square)

# Goal: clean product shot, white background
"a single blue wireless headphone floating on pure white background
product photography style, centered composition
clean white studio background
NO shadows visible, NO reflections, NO background objects
NO models or people, NO props
professional product shot, square format"
size: "1024x1024"
Combined formula: subject description + composition/ratio keywords + style + NO exclusion list

07 | Summary

Two techniques in one sentence:

Negative constraints = use NO / without / avoid to tell the AI what not to generate.

Aspect‑ratio control = choose the appropriate size value and add composition keywords that match your use case.

Best practices:

Place exclusion terms at the end of the prompt.

Use uppercase NO for stronger effect.

Apply three‑layer exclusion (element → style → scene) as needed.

Pick size based on the intended placement, not just visual preference.

Portrait cover + landscape illustration is the golden combo for public‑account content.

Original Source

Signed-in readers can open the original source through BestHub's protected redirect.

Sign in to view source
Republication Notice

This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactadmin@besthub.devand we will review it promptly.

Prompt EngineeringAI image generationaspect ratioGPT Image 2negative prompts
James' Growth Diary
Written by

James' Growth Diary

I am James, focusing on AI Agent learning and growth. I continuously update two series: “AI Agent Mastery Path,” which systematically outlines core theories and practices of agents, and “Claude Code Design Philosophy,” which deeply analyzes the design thinking behind top AI tools. Helping you build a solid foundation in the AI era.

0 followers
Reader feedback

How this landed with the community

Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.