OpenAI just announced API access to its new and widely improved, powerful AI image generator, DALL-E 3. This latest version promises even higher-resolution images and more realistic and nuanced outputs than its predecessor.
What It Includes
- Higher resolution outputs and more realistic details. The API now enables 1024x1024 images as well as larger 1792x1024 sizes.
- One major update is the addition of an hd quality setting, which can provide enhanced image quality and crisper details. This comes at the cost of longer generation times compared to the standard quality.
- Includes built-in content moderation to filter inappropriate or harmful outputs, helping developers avoid misuse
- While DALL-E 2 allowed up to 10 images to be generated per prompt, the DALL-E 3 API limits it to 1 image at a time, likely due to the increased processing power required. Although developers can make parallel requests for multiple images if they want to get around this.
- The API also provides options beyond generating from scratch, like creating variations of existing images and editing parts of images based on textual prompts. This could unlock creative workflows and new visual effects.
Early partners like Snap, Coca-Cola and Shutterstock have already been utilizing DALL-E 3 to create images for campaigns and customers. While visual AI has sparked some concerns around misuse, OpenAI says this new API will have the same built-in content moderation as previous versions to help developers prevent abuse.
Pricing for the API starts at 4 cents per image, putting high-end AI image generation well within reach of startups and developers. I'm sure this price will decrease given time too. This is just the start.
This also comes a few minutes after OpenAI announces a promise to defend its API and enterprise customers against copyright complaints.
Overall, the DALL-E 3 API is way more customizable and complex than its predecessor. There's a lot more you can do and a lot more doors to be opened.