DALL·E: image generation that follows instructions
DALL·E's strength is literal prompts and short text in images, which makes it useful for diagrams and simple assets.
Pricing
Included with ChatGPT paid plans; API billed per image (check the official page)
Official site
Open the official website (opens in a new tab)
Prices and plans change. Check the official page before you pay.
DALL·E is OpenAI’s image generator, available inside ChatGPT and through the API. Its distinguishing trait is obedience: describe a scene precisely and it usually places the elements where you asked.
What it is good at
- Literal prompts. "A white ceramic mug on a wooden desk, a notebook to the left, morning light from a window on the right" produces something close to that.
- Short text in images. Signs and labels are often spelled correctly, which most generators still struggle with.
- Simple illustrations and icons for a slide or a draft post.
- Editing by conversation. "Make the background lighter and remove the chair" is a normal follow-up inside ChatGPT.
Where it fails
Fine detail, hands, and anything requiring consistent characters across images. It also has a strong default style that can make every image look like the same illustrator; asking for a medium helps ("flat vector", "watercolour", "photograph with a 50mm lens").
It refuses some categories of content, and the refusal can be surprising when a prompt is ambiguous. Rewriting the request is usually faster than arguing with it.
Prompting
DALL·E responds to sentences rather than keyword lists. A useful structure:
- Subject and action.
- Setting and background.
- Light and time of day.
- Medium and style.
- Composition or aspect ratio.
Keep the request to one scene. "A market with five stalls, each with different produce" tends to produce a generic market; five separate images work better.
Pricing
Image generation is included in ChatGPT’s paid plans with limits, and the API charges per image at rates that depend on size and quality. Check the official pricing page; the numbers change with model versions.
Licensing and disclosure
Images can be used commercially under OpenAI’s terms, and the output may not be protectable by copyright. Several platforms require disclosure of AI-generated media, and some publications do not accept it at all. Check the policy of wherever the image will appear.
Who it is for
Anyone who needs a quick, specific illustration: a blog header, a slide background, a mockup. It is not for product photography, technical diagrams with exact labels, or anything where a plausible-looking error would mislead.
Alternatives
Midjourney produces more striking, styled images and less literal ones. Adobe Firefly focuses on commercially safe training data and Photoshop integration. Stable Diffusion is the choice when you need control or local generation.