ChatGPT now supports DALL·E 3. While still in beta, it’s capable of generating high-quality images from concise prompts.
Usage Guide
The free tier of ChatGPT doesn’t include access to the DALL·E 3 model. To utilize it, you’ll need a “ChatGPT Plus” subscription.
Hover over the New Chat Button, and a model selection option will pop up. Choose “DALL·E 3” from the dropdown.
Do note that since it’s a beta feature, availability may vary. I once noticed an update saying “DALL·E 3 available” on my X (Twitter) timeline, but upon checking my account, it hadn’t shown up yet.
To generate an image, simply type your prompt into the chat field. The system supports prompts in multiple languages. For instance, I used Japanese for my test.
Generating Chihuahua Images
I input the following prompt.
高級車の窓から顔を出し、風に吹かれて、気持ちよさそうにしている、ロングコートチワワ – A long-coated Chihuahua, peeking its face out of a luxury car window, hair blowing in the wind, appearing utterly relaxed.
Within 20 seconds, these four images were generated.
For comparison, I entered the same prompt into DALL·E 2. The resulting images were somewhat darker. DALL·E 3 outperforms its predecessor.
Context Retention
A remarkable feature of DALL·E 3’s integration with ChatGPT is its ability to retain context. This allows users to provide follow-up prompts based on previous results. For instance, I entered:
もう少し毛が短い感じにしてください。 – Please make the fur appear slightly shorter.
The updated images generated were:
DALL·E 3 seamlessly interpreted my feedback, producing a natural-looking result. The fur appeared naturally trimmed. One illustration even resembled my pet Chihuahua, Lala, right after her grooming session.
Testing with Brief Prompts
I also experimented with shorter, more direct prompts:
ロングコートチワワ – A long-coated Chihuahua
The images generated were:
The clarity and quality were astonishing, given the brevity of the prompt.
DALL·E 3 has made significant advancements.
