DALL-E 3

Overview

DALL-E 3 by OpenAI was a breakthrough image model, now succeeded by GPT Image 1 but still widely used. Known for excellent text rendering in images and strong instruction following.

Key Strengths

• Text in images: Best-in-class text rendering (signs, labels, typography)
• Instruction following: Accurately follows complex compositional instructions
• Safety: Strong content policy with automatic prompt revision
• Consistency: Reliable output quality across diverse prompts
• Integration: Native in ChatGPT, easy API access

Prompting Tips

Be Descriptive, Not Technical

DALL-E 3 prefers natural language over technical photography terms:

GOOD: "A cozy coffee shop interior with warm morning light streaming through large windows, a barista making latte art, plants on shelves, exposed brick walls"
LESS EFFECTIVE: "Interior shot, 35mm f/1.4, golden hour, bokeh, shallow DOF"

Specify Style Explicitly

"...in the style of a watercolor painting"
"...as a minimalist vector illustration"
"...photorealistic, shot on a DSLR camera"
"...in the style of a 1950s advertisement"

Text on Images

DALL-E 3's killer feature — use it:

"A book cover with the title 'MIDNIGHT GARDEN' in elegant serif font, author name 'Sarah Chen' below"

Aspect Ratios

• Square: 1024x1024 (default)
• Landscape: 1792x1024
• Portrait: 1024x1792

Best Use Cases

Use Case	Quality
Text/typography in images	Excellent
Product mockups with labels	Excellent
Book covers & posters	Very Good
Illustrations & diagrams	Very Good
Photorealistic scenes	Good
Character consistency	Fair

vs GPT Image 1

GPT Image 1 (successor) has better photorealism and more precise control, but DALL-E 3 is still better at text rendering and more predictable.

Limitations

• Automatic prompt revision (ChatGPT rewrites your prompt — can change intent)
• No negative prompts
• No img2img or inpainting (GPT Image 1 added this)
• Style consistency across multiple images is difficult