OpenAI has introduced an enhanced iteration of its text-to-image tool, known as DALL-E. This new version leverages ChatGPT, OpenAI’s widely recognized AI chatbot, to simplify generating image prompts.
Contemporary AI-powered image generation tools translate image descriptions or prompts into diverse artistic styles, from highly realistic to fantastical renditions. However, formulating the perfect prompt often poses a significant challenge, to the extent that “prompt engineering” has emerged as a dedicated profession.
OpenAI’s latest offering, DALL-E 3, integrates ChatGPT to assist in refining prompts. Subscribers of OpenAI’s premium ChatGPT plans, ChatGPT Plus, and ChatGPT Enterprise, can now enter an image request and fine-tune it through interactions with the chatbot, receiving the results directly within the chat application.
ChatGPT can enhance prompts, even with just a few words, by making them more descriptive and offering additional guidance to the DALL-E 3 model.
The integration of ChatGPT is not the sole enhancement featured in DALL-E 3. OpenAI reports that DALL-E 3 generates higher-quality images, particularly when handling longer prompts, which closely align with the intended descriptions.
Moreover, it exhibits improved performance when dealing with content that has historically posed challenges for image-generating models, such as textual descriptions and depictions of human hands.
Beyond these improvements, DALL-E 3 incorporates novel mechanisms to mitigate algorithmic bias and enhance safety. For instance, DALL-E 3 will reject requests that seek images resembling the work of living artists or featuring public figures.
Additionally, artists now have the option to exclude specific pieces or all of their artwork from being used for training future iterations of OpenAI’s text-to-image models. (OpenAI and some of its competitors face legal scrutiny for allegedly utilizing copyrighted artwork by artists to train its generative AI image models.)
The launch of DALL-E 3 occurs amidst intensifying competition in the generative AI field, particularly in image synthesis. Competitors like Midjourney and Stability AI continue to refine their image-generating models, increasing the competitive pressure on OpenAI to maintain its leadership.
OpenAI’s plan initially involves introducing DALL-E 3 to premium ChatGPT users in October, making it accessible to research laboratories and API customers. However, the company has not disclosed whether it intends to release a free web tool similar to its previous offerings, DALL-E 2 and the original DALL-E model. It has left such a release with uncertain timing.