OpenAI Announces the Release of DALL-E 3

This week OpenAI announced the upcoming release of DALL-E 3, their advanced text-to-image system; stating that it offers a big improvement over its predecessors in translating nuanced and detailed text into highly accurate images. It is integrated with ChatGPT, ensuring enhanced user interaction and creative control, and is set to be available for ChatGPT Plus and Enterprise customers in October.

?

Summary of Announcement

• Integration with ChatGPT: DALL-E 3 is built natively on ChatGPT, allowing users to use ChatGPT for refining and brainstorming their prompts and ideas. This may finally provide DALL-E with a large leap in adoption/usage, since it will be so readily available to ChatGPT’s large user base.
• Enhanced User Interaction: Users will be able to use natural language to create images from their ideas, ask for modifications, and have creative discussions, purportedly making the tool highly user-friendly and interactive.
• Availability: It will be available to ChatGPT Plus and Enterprise customers in October and will be accessible via the API and in Labs later this fall.
• Creative Control and Usage Rights: Users have the rights to the images they create using DALL-E 3 and can reprint, sell, or merchandise them without needing permission from OpenAI. The system also allows creators to opt their images out from training of future image generation models.
• Refinement in Image Generation: DALL-E 3 will also reportedly be able to generate images that adhere exactly to the provided text, overcoming the limitations of previous models that tended to ignore certain words or descriptions.

?

HOT TAKE:

Earlier iterations of DALL-E have been underwhelming when compared with Midjourney, Stable Diffusion, and Leonardo.ai (and others). Image quality has not been comparable and the ability to manipulate images (panning, zooming, in-painting, etc.) lacking. It has yet to be seen just how big an improvement DALL-E 3 will be, but the accessibility from the integration with ChatGPT (not to mention a more user friendly natural language interface), will be a huge adoption engine if the quality is anywhere near that of its competitors. Midjourney (with its Discord only access) and Stable Diffusion (with it’s offline PC version), are much less user friendly, even with their significantly advanced abilities. I suspect they will most likely remain the go-to option for advanced users and special cases, with zoom features, in-painting and the ability to use LoRAs and a plethora of models and integrated tools. One other potential DALL-E 3 win would be the ability to integrate text which is here-to-for problematic in other models.

#dall-e3, #diffusionmodels, #chatgpt, #imagegeneration, #openai