If you already know the basics of prompting, come join us to level up your skills with DALL·E 3.
In this tutorial, you’ll learn advanced techniques and creative tricks to generate stunning images with precision and style.
Section 0: Getting Started — Logging In and Accessing DALL·E
What is DALL·E 3?
DALL·E 3 is the latest image generation model by OpenAI, integrated directly into ChatGPT. It allows users to generate and edit images using natural language prompts.
Where to access it:
The most up-to-date version of DALL·E is available inside ChatGPT (https://chat.openai.com) for users with a ChatGPT Plus subscription.
Steps to log in and start generating images:
- Go to https://chat.openai.com
- Sign up or log in to your OpenAI account.
- Click your name or profile icon in the bottom left and go to Settings.
- Under “My Plan”, upgrade to ChatGPT Plus ($20/month).
- In the model switcher at the top, select GPT-4 (you’ll be using GPT-4o, which includes DALL·E 3).
- You’re now ready to generate images by typing prompts into the chat.
To generate an image, simply enter a prompt like this:
“Create an image of a castle floating in the clouds, digital painting style.”
ChatGPT will return the generated image directly in the conversation.
Section 1: The Fundamentals — How DALL·E Thinks
DALL·E is a language-to-image model. It doesn’t “see” or “imagine” like a human. It creates images by predicting visual outcomes based on text descriptions. That means it responds best to clear, descriptive language—especially language grounded in visual, artistic, and emotional cues.
To get great results, you must describe your scene as though explaining it to a professional illustrator or cinematographer.
Section 2: Structuring a Strong Image Prompt
To reliably control the results, write prompts that include specific visual attributes. The more you define the visual world, the less randomness DALL·E introduces.
Use this structure for your prompts:
[Subject] doing [Action], in [Setting/Environment], at [Time of Day], with [Lighting and Color], in the style of [Artist or Medium], conveying a [Mood or Emotion]
Example:
“A woman standing in a wheat field at sunset, with warm golden light, soft shadows, and a calm expression, in the style of an oil painting, evoking peace and nostalgia.”
Section 3: Visual Thinking — Moving from Concept to Prompt
Before writing, ask:
- What is the subject?
- What is the setting or environment?
- What emotion or atmosphere do I want to create?
- What style or medium should it look like?
- What kind of lighting and color palette would match?
Example Concept
Idea: “A sense of isolation in a futuristic world”
Step-by-step translation:
- Subject: A lone figure
- Environment: Dystopian cityscape
- Mood: Isolation
- Style: Sci-fi concept art
- Time and light: Night, glowing neon
Prompt:
“A lone figure walking down an empty neon-lit street in a futuristic city, deep shadows and glowing signs, night scene, cinematic sci-fi style, inspired by Blade Runner.”
This prompt tells DALL·E what to show, how it should feel, and what style to emulate.
Section 4: Controlling Style
DALL·E supports a wide range of styles. Naming a style in your prompt helps guide the composition, color, and texture of the image.
Common style terms you can use:
- Oil painting
- Watercolor
- Ink drawing
- Digital illustration
- 3D render
- Pixel art
- Retro comic book
- Cyberpunk
- Ukiyo-e (Japanese woodblock)
- Surrealism
- Art deco
- Minimalist line art
Example prompt:
“A giraffe wearing sunglasses walking through Times Square, in the style of a 1990s comic book cover.”
Why it works:
You’ve described the scene and given a visual reference style, which DALL·E can map to a specific type of color, texture, and composition.
Section 5: Iterating and Refining Prompts
Creating high-quality images with DALL·E is an iterative process. Your first image is a sketch. Each following prompts should refine or evolve it.
Start simple:
“A dragon flying over mountains.”
Add detail:
“A golden dragon flying above snow-covered mountains under a twilight sky, glowing clouds, fantasy illustration.”
Refine style and mood:
“A golden dragon soaring above icy mountains at dusk, wings reflecting orange light, cinematic fantasy art in the style of Magic: The Gathering card illustrations.”
Each refinement clarifies:
- Subject appearance
- Lighting
- Emotion
- Visual format
Section 6: Using Inpainting to Edit Images
DALL·E 3 (via ChatGPT Plus) supports inpainting, which allows you to change part of an image after it has been generated.
How to use inpainting:
- Generate an image with a prompt.
- Hover over the image in ChatGPT and click the “Edit” (pencil) icon.
- Use the brush to select the area you want to modify.
- Enter an instruction, such as “Replace the tree with a small pagoda.”
- DALL·E will regenerate just the selected area.
This is ideal for:
- Adding or removing elements
- Correcting details (lighting, faces, objects)
- Evolving a scene step-by-step
Section 7: Creating Custom Visual Styles
You can direct DALL·E to create unique visual styles by blending influences.
Example prompt:
“A cityscape rendered in a hybrid style combining Japanese ukiyo-e woodblock art and modern architectural sketching, monochrome with red accents.”
This kind of prompt works because it:
- Names two style references
- Specifies the medium and texture (woodblock + sketch)
- Introduces a color limitation (monochrome + red)
You can blend:
- Historical art styles with sci-fi
- Real-world cultural references with digital formats
- Natural textures with abstract design
Section 8: Composition and Directional Language
To further control the image, include composition cues. DALL·E understands basic photographic and cinematic language.
Terms to use:
- Wide shot / close-up
- Bird’s-eye view / top-down / isometric
- Symmetrical / centered / off-center
- Portrait orientation / landscape orientation
- Backlighting / front lighting / rim light / ambient light
- Mood words: eerie, joyful, tranquil, apocalyptic
Example prompt with composition:
“A child looking out a window at a rainy cityscape, viewed from behind, soft diffused lighting, shallow depth of field, photographic realism.”
Here, you’re directing the perspective, lighting, and style.
Section 9: Prompt Templates for Common Use Cases
Character Portrait
“A [type of person or creature], wearing [style or clothing], in [pose or expression], background of [environment], in the style of [artist or medium], with [lighting and mood].”
Example:
“A medieval knight in weathered armor, standing in profile against a stormy battlefield, realistic digital painting with dramatic lighting.”
Landscape
“A [landscape or environment], under [time of day and weather], seen from [angle], in the style of [painting or media], conveying a sense of [emotion or scale].”
Example:
“A vast desert at sunset, viewed from a high dune, shadows stretching far, soft orange and purple tones, in the style of a watercolor painting, evoking loneliness.”
Surreal Concept
“A [subject or object] in a world where [unusual twist], rendered in [style or medium], with [color palette and lighting], inspired by [surrealist artist or film].”
Example:
“An elephant made of clock gears walking across a frozen ocean, in the style of Salvador Dalí, with melting shadows and surreal lighting.”
Section 10: Ethical Use and Exporting Images
Saving and using images:
- Right-click on generated images to save.
- You are permitted to use DALL·E images for personal and commercial use under OpenAI’s terms.
- Avoid using generated images to impersonate real individuals or mislead others about authorship.
Things to avoid:
- Generating nudity, gore, or harmful content
- Trying to recreate real people’s likenesses
- Misusing AI art in misleading advertising or journalistic contexts
Section 11: Practice Challenge – Evolve a Prompt in 3 Steps
Starting prompt:
“A lighthouse on a cliff”
Refinement 1:
“A white lighthouse on a jagged cliff at sunset, waves crashing below, digital painting.”
Refinement 2:
“A towering white lighthouse on a crumbling cliff during a storm, lightning striking in the background, dramatic shadows, painted in 19th-century Romantic style.”
Refinement 3:
“A surreal scene of a lighthouse floating above the ocean, beams of light piercing the sky, oil on canvas style with bold brushstrokes, dreamlike atmosphere.”
This practice teaches:
- Prompt layering
- Mood shaping
- Style transitions
- Creative remixing
Summary
AI image generation is a creative process that rewards visual thinking, iteration, and control. To become proficient with DALL·E, you need to:
- Write descriptive, structured prompts
- Direct style, lighting, and composition
- Iterate and refine results step-by-step
- Use inpainting to edit specific areas
- Explore and invent custom styles
- Think like a visual storyteller
