OPENAI INTEGRATES DALL-E 3 WITH CHATGPT: EVERYTHING THAT’S GOING TO CHANGE

OpenAI has recently announced the third version of its text-to-image generator platform DALL-E.

What's new is it's integrated with ChatGPT for easier prompt generation and includes more safety options. The DALL-E integrated ChatGPT API will be available to ChatGPT Plus and Enterprise subscribers starting in October.

What’s Changed?

Dall-E 2 was capable of generating marvellous and captivating artwork. However, it lacked in some areas.

Some of the shortcomings of Dall-E 2 are:

Limited understanding of context

Dall-E 2 could not understand the contextual relationship between different objects. As a result, the final output lacked quality and naturalism.

Difficulty with complex images

Though Dall-E 2 came up with best-in-class upgrades, it still lacked to process detailed prompts. The system could not grasp the intricacies of the prompts and failed to generate detailed images.

If you ask Dall-E 2 to produce an image with multiple characters, it will surely miss out on the details. Moreover, it was a disappointment to see Dall-E 2 struggling to process complex descriptions.

Mistakes in composition

Many times I have observed that Dall-E 2 mixes the composition of the characters or objects. It is because the system could not understand the relation between the characters and ended up wrongly placing them in the image.

Need for detailed input

One of the major drawbacks of Dall-E 2 was its inability to generate quality images from simple and short prompts.

You will have to enter detailed prompts with very specific briefs about how you want the tone, layout and colours. Now not everyone can pen down their vision into detailed paragraphs.

If your description is vague or unclear, the resulting image may not be up to your liking.

Potential to generate harmful images

With the world promoting the ethical use of AI, Dall-E 2 had a major drawback here.

The model was not trained to avoid harmful content generation. It could sometimes generate images that were biased, racial and offensive.

?

So, the OpenAI team put in a lot of work to enhance the capabilities of Dall-E 3. They ensured that the new version stands true to its quality, compliance and security standards.

The development team at OpenAI has supercharged Dall-E 3 to:

·?????? Include more safety standards

·?????? ChatGPT assistance in prompt generation

·?????? Enhance context understanding

·?????? Generate detailed images

·?????? Improve image quality

·?????? Coherent text incorporation

·?????? Ensure responsible use of AI

·?????? Robust safety measures

·?????? Grasp the context of prompts

·?????? Equipped to understand the details you provide

These upgrades will enable Dall-E 3 to produce exemplary images just the way you want or even better.

What we did with Dall-E 2 was we had to write detailed prompts to create images that aligned with our vision. You have to be very specific about how you want your image to be. Otherwise, the model is going to mess up.

Now the problem is that:

·?????? Not everyone is good at writing detailed prompts.

·?????? Sometimes we cannot explain in words what we imagine.

·?????? You may miss the little details in the prompts.

·?????? Writing several detailed prompts to finalise one image takes a lot of time.

·?????? Even if you write good prompts, the model may not understand the composability and context.

·?????? So, you end up having a low quality image.

·?????? The whole process is time-consuming and tiring.

Dall-E 3 resolves the prompt issue with ChatGPT integration. Now you don’t have to write detailed prompts. Rather ChatGPT can help you generate prompts for Dall-E.

All you need to do is enter the prompt ChatGPT generated and get creative images. ChatGPT will write paragraphs with longer sentences in a way that Dall-E 3 easily understands.

This ensures that the model accurately grasps the context and details of the prompt resulting in exceptionally magnificent images.

In the demo to The Verge, Aditya Ramesh asked ChatGPT to write a prompt for the logo of a ramen restaurant. The ChatGPT then came up with 4 options.

The most creative one was where the ChatGPT generated a prompt describing mountains with snowcapped ramen, broth flowing like a river and pickled eggs displayed as stones.

Though this was more of an illustration and art rather than a logo, it still raises the bar of creativity for AI generated images.

?

Will DALL-E 3 Be A Responsible AI?

To promote the responsible use of AI, the OpenAI team has put in place rigorous safety measures.

The model will now prevent the generation of lewd, offensive and potentially harmful images. Further, the model will avoid recreating images resembling public figures keeping in focus the ethical standards and privacy.

That's not all! You cannot create art similar in style and tone to another artist. This is to protect the authenticity and creativity of artists.

What about Copyright rules?

In the past artists have filed lawsuits against, DALL-E, Midjourney and Stability AI for using their art without consent for training and other purposes.

To avoid expensive and controversial lawsuits, OpenAI has released opt-out feature for the artists.

They can now opt out of reusing their AI generated art for training future image generation models.

Creators can submit an image that they own the rights to and request its removal in a form on its website. A future version of DALL-E can then block results that look similar to the artist’s image and style.

These steps will redefine the AI generated art landscape laying the foundation for ethical AI use in the art.

?

To sum up, we are already very excited about the upcoming Dall-E 3 ChatGPT integration. With unmatched artistic capabilities combined with simplified prompt generation and safety features, Dall-E 3 will be leading the world of generative AI art.

AI assisted creativity never looked so promising.

?

?

If you are into AI, LLMs, Digital Transformation and Tech world – do follow me on LinkedIn.

Stay tuned for my insightful articles on every Monday.

?

?