Image Generation & Analysis

ChatGPT uses DALL-E to turn your text prompts into images. You simply describe what you want, and the AI generates it. No design skills needed.

The key to great image generation is specificity. Compare these two prompts:

Vague: "A dog in a park"

Specific: "A golden retriever sitting in a sunlit autumn park, fallen orange and red leaves scattered on the ground, soft bokeh background, warm afternoon light, photorealistic style"

The more detail you provide about subject, setting, lighting, style, and composition, the better your results will be. Think of yourself as an art director giving instructions to a designer.

Here are the style categories you can use:

Photorealistic — looks like a photograph, great for product mockups and stock imagery

Illustration — hand-drawn, digital art, or vector styles for blog posts and social media

3D render — clean, modern 3D objects for tech presentations and app visuals

Watercolor / Oil painting — artistic styles for creative projects and wall art

Flat design / Minimalist — clean icons and simple visuals for UI and branding

ChatGPT's vision capabilities let you upload any image and have a conversation about it. This is incredibly powerful for real-world tasks.

What you can upload and analyze:

Screenshots — paste a screenshot of an error message, a website design, or a dashboard, and ChatGPT will explain what it sees and offer suggestions.

Diagrams and charts — upload a flowchart, org chart, or data visualization. ChatGPT can read it, explain it, and even recreate it in a different format.

Photos of documents — snap a picture of a whiteboard, handwritten notes, or a printed document. ChatGPT will transcribe and organize the content.

Product photos — upload images for analysis, comparison, or to generate marketing copy based on what it sees.

Design mockups — share wireframes or UI designs and get feedback on layout, usability, and accessibility.

To upload an image, click the attachment icon (paperclip) in the ChatGPT message bar, or simply paste an image from your clipboard.

One of the most powerful features is image editing through conversation. You can ask ChatGPT to modify generated images or even work with images you upload.

Here are practical editing workflows:

Iterative refinement — Generate an image, then ask for specific changes. "Move the text to the left." "Make the background darker." "Add a subtle drop shadow." Each instruction builds on the previous result.

Style transfer — Take a concept and change its visual style. "Now make this same scene as a pixel art illustration." "Convert this to a vintage poster style."

Element addition/removal — "Remove the person in the background." "Add a coffee cup on the desk." "Replace the sky with a sunset."

Text and branding — "Add the text 'Launch Day' in a bold sans-serif font at the top." "Place our logo in the bottom-right corner."

The editing workflow is conversational. You don't need to start over — just keep refining until you get exactly what you want.

Here's a formula for writing effective image generation prompts. Include as many of these elements as relevant:

1. Subject — What is the main focus? ("A woman working on a laptop", "A futuristic city skyline")

2. Setting/Environment — Where is it? ("in a cozy cafe", "on a rooftop at sunset")

3. Style — What visual style? ("photorealistic", "watercolor illustration", "flat vector design", "cinematic")

4. Lighting — What's the mood? ("soft natural light", "dramatic side lighting", "neon glow")

5. Composition — How is it framed? ("close-up portrait", "wide-angle aerial view", "centered symmetrical")

6. Color palette — Any color preferences? ("warm earth tones", "monochrome blue", "vibrant and saturated")

7. Mood/Atmosphere — What feeling? ("peaceful and calm", "energetic and dynamic", "mysterious and moody")

Pro tip: You can also use reference styles like "in the style of a New Yorker magazine illustration" or "like a Studio Ghibli background painting" to guide the aesthetic.