How to use DALL-E 3?
As of July 2026 DALL-E 3 is retired, so you use its successor the same way: open ChatGPT at chatgpt.com, sign in, and type an image request in plain English like 'create a watercolor fox in a forest.' ChatGPT generates it with GPT Image. It's free with a small daily cap, or unlimited-feeling on ChatGPT Plus ($20/month).
Why — the first-principles explanation
Using DALL-E 3 was never about learning software; it's about describing what you want in words. That design choice, natural-language prompting, is why it took off: the model reads your sentence, converts it to numbers, and paints a matching image. The clearer and more specific your description, the closer the result.
Since December 2025, ChatGPT's image button no longer calls DALL-E 3, it calls GPT Image, DALL-E's successor. For you, the workflow is identical: you still just chat. You can also iterate in conversation: after the first image, say "make it nighttime" or "add a red scarf," and it edits or regenerates. This conversational refinement is the core skill, treat it like directing an artist, not filling out a form.
Good prompts name the subject, style, setting, mood, and details: "a cozy cabin at dusk, oil-painting style, warm glowing windows, snow falling, wide shot." Vague prompts get generic results. You can request specific aspect ratios or ask for text in the image, GPT Image renders words far better than older models. When you hit the free daily cap, ChatGPT tells you when it resets or offers an upgrade.
An example that makes it click
It's like ordering a custom cake by talking to a baker. You don't operate any machines; you just say, 'a two-tier chocolate cake, gold trim, with a tiny fondant rocket on top.' If it's not quite right, you say, 'make the rocket bigger and add stars.' The baker adjusts. DALL-E 3, now GPT Image, is that baker for pictures, and your words are the whole recipe.
How to do it
- Go to chatgpt.com and sign in, or create a free account.
- Start a new chat and type a clear image request, naming subject, style, setting, and mood.
- Wait a few seconds for the image, then refine it in the same chat ('make it nighttime', 'add a scarf').
- Download the final image with the save icon.
- If you hit the free limit, note the reset time or upgrade to ChatGPT Plus for more generations.
Key facts
- DALL-E 3 was retired in 2026; ChatGPT now generates images with GPT Image using the same chat workflow.
- You create images by typing natural-language prompts, no separate software needed.
- You can refine images conversationally by asking for changes in follow-up messages.
- Specific prompts (subject, style, setting, mood, details) produce better results than vague ones.
- Free use has a small daily cap; ChatGPT Plus ($20/month) raises it.
OpenAI's image generator, built into ChatGPT.
Affiliate link — we may earn a commission at no cost to you.▶ The 60-second explainer (script)
Here's how to use DALL-E 3, keeping in mind it's now been upgraded to GPT Image inside ChatGPT. It's simple: you just describe what you want. Go to chatgpt.com, sign in, and type something like 'create a watercolor fox in a forest.' A few seconds later, your image appears. The real trick is being specific, name the subject, the style, the setting, and the mood. So instead of 'a house,' try 'a cozy cabin at dusk, oil-painting style, warm glowing windows, snow falling.' Even better, you can refine by chatting: say 'make it nighttime' or 'add a red scarf,' and it adjusts, like directing an artist. Free accounts get a couple of images a day; ChatGPT Plus at twenty dollars a month gives you plenty more. That's it, no software to learn, just words.
What authoritative sources say
People also ask
Do I need any special app to use DALL-E 3?
No. You use it through ChatGPT in a browser or the ChatGPT app. You just type what you want; there's no separate DALL-E software to install.
How do I write a good prompt?
Be specific: name the subject, art style, setting, mood, lighting, and any details. Vague prompts give generic images; detailed prompts give what you pictured.
Can I edit an image after it's made?
Yes. In the same chat, ask for changes like 'make it brighter' or 'add a hat,' and the model regenerates or edits accordingly.
Is it still called DALL-E 3?
The feature is the same, but the model is now GPT Image, DALL-E 3's successor. Many guides still say DALL-E 3 out of habit.