[api设置] Introduction to the DALL·E 3 API and how to use the DALL·E 3 API

What is it?
DALL·E 3 is an image generation model created by OpenAI. The DALL·E 3 API can:

Generate text in images
Support landscape and portrait images
Generate more detailed images
Understand complex prompts

Note: Because DALL·E 3 requires very detailed prompts, the API will automatically create more detailed prompts, just like in ChatGPT.

How do I access it?
Anyone with an OpenAI API account can access the OpenAI API. Use the “dall-e-3” model parameter to access the new model. ChatGPT Plus subscribers can also access DALL·E 3 through ChatGPT.

Is DALL-E 2 still available?
Yes, the DALL·E 2 API will still be available. For backward compatibility reasons, the API will default to DALL·E 2, but you can set the model parameter to dall-e-3 to switch to that model.

Does DALL-E 3 have different rate limits? How do I increase these limits?
You can view DALL·E rate limits in the rate limits guide.

Why can I no longer generate 512 or 256 images?
DALL·E 3 is trained to generate 1024x1024, 1024x1792, or 1792x1024 images. To create images faster or reduce quality and cost, you can use the new parameter quality. If you choose the dall-e-2 model, you can still generate 512x512 or 256x256 sized images.

What is the difference between the quality options?
The quality attribute is available when using DALL·E 3. The default is “standard,” which will create images quickly and at lower cost. Users can specify quality: "hd" (and pay a higher price) to give the model more time to generate images, resulting in higher image quality but also higher latency.

What do the style options do?
OpenAI introduced a new optional parameter, style, to provide advanced control over generation. Currently, the valid options are “vivid” or “natural.” OpenAI recommends trying this new parameter to determine which option best suits your use case. The default is “vivid.”

Why does DALL·E-3 only support n=1?
For reasons of system scalability and reliability, OpenAI currently only supports n=1 when calling DALL·E 3. If you want to receive more than 1 image, OpenAI recommends calling the API multiple times in parallel.