OPENAI reveals Dall-E 3, an image generator based on Chatgpt, Dall-E 3: Fusion between chatgpt and the generation of images that will shake up Midjourney

Dall-E 3: The fusion between Chatgpt and the generation of images that will shake up Midjourney

By integrating Dall-E at Chatgpt, OpenAi strengthens its position as a leader in the generative AI world. Chatgpt and Dall-E 3, together, promise to repel the borders of AI.

OPENAI reveals Dall-E 3, an image generator based on Chatgpt

The latest version of the model is based on GPT to facilitate the creation of complex and coherent images.

OPENAI, the company behind the essential GPT, has just revealed its latest generative model to date. And this time, it is again the image that is in the spotlight; Sam Altman’s troops revealed the third version of the Dall-E image generator. He promises more shades, details and complexity.

The first point highlighted by Openai is the accuracy of prompt. Even the best generators of the moment, such as Midjourney or Dreamstudio, suffer from a common problem: they have tendency to ignore certain words of the textual request.

To get the best results, it is therefore necessary Empirically learn to communicate with algorithm. This implies making lots of tests on vocabulary and turns to find those who are most likely to be interpreted correctly. This is an exercise that can be delicate when the user awaits a very specific result, hence the term ” prompt engineering »».

Chatgpt mobilized for the composition of prompt

With Dall-E 3, Openai wants to end this state of affairs. “” Modern text-to-image systems tend to ignore words or descriptions, which forms the user to learn the promineering prompt. Dall-E 3 represents a big step forward in our ability to generate images that adhere exactly to the text provided “Indicates the company.

And to do this, the company already has the perfect tool: Chatgpt. The new version of the Diffusion 3 model is ” Natively built »Around the chatbot. It now serves as an intermediary for Interpret every detail of the request. The language model can also help the user generate prompt extremely detailed so that Dall-E understands exactly what is expected of him without using techniques of prompt engineering advanced.

And if the initial result is not suitable, that it does not. It is enough to go again by chatgpt to claim precise changes in terms of color, shape, style, a particular detail … In other words, Openai has put the small dishes in the big ones to offer A maximum control degree with A minimum of effort and technical expertise.

On the examples proposed by Openai, also note that Dall-E 3 is able to generate coherent texts. This is something that image generators had a lot of trouble doing so far. For example, the latest version of Midjourney always produces enough … exotic results when asked to engage in this exercise.

Protect public figures

The company still imposed a number of limits on this third version. For example, engineers have consolidated the safeguards already present on Dall-E and Dall-E 2 which prevent the user from generating content ” violent, adult or hateful »».

In addition, Dall-E 3 is designed to refuse requests that explicitly mention a public figure. The goal is to prevent it from being used for propaganda or disinformation purposes.

Respect intellectual property

The other important point of this update is that Optai seems to have taken note of the many criticisms that have been addressed to the various image generators about intellectual property. Dall-e goes refuse the requests that ask him to imitate the style of a living artist. This should close the door to numerous counterfeit attempts.

On the other hand, this protection system will most certainly be imperfect. The precision of this third version is a double -edged sword. In theory, a sufficiently perceptive user can always obtain a fairly close result. You will just have to be able to describe the style of the targeted artist very precisely without explicitly name.

Openai also offers two solutions that allow an artist of refuse that his work is used to train Dall-E. The first is to fill out the form at this address. But if you are looking to protect a large number, like the ones you keep on your online portfolio, the ideal is to prohibit access to the Crawler GPTBOT. To do this, just follow the instructions available at this address.

This will considerably limit the ability of Dall-E 3 to imitate the works in question. At least, if they have not already been passed to the reel. And it is unfortunately impossible to determine in practice, since the source data is not kept after being integrated into the model.

Dall-E3 will be officially launched next October for customers of Chatgpt Plus and Chatgpt Entreprise offers.

Dall-E 3: The fusion between Chatgpt and the generation of images that will shake up Midjourney

OPENAI, the mother’s parent company, is no longer to be presented. After having offered us Chatgpt, she now announces the latter’s ability to generate precise images, thanks to the fusion of Dall-E and Chatgpt in Dall-E.

Frandroid, For those who follow, use an image generation tool: Midjourney. If this instrument turned out to be powerful and innovative in the world of image generation, Dall-E of Openai was also on the track, with an equally impressive course. And the new opus Dall-e 3 Risk well to change the situation.

According to the demonstrations shared by Openai, it is clear that Dall-E 3 is distinguished from its predecessors and competitors. Where Midjourney excels in photorealistic restitution, but still requires many manipulations, Dall-e 3 promises a much more efficient and less erroneous image generation. A notable feat, especially considering the fact that this is a major advance compared to Dall-E.

In addition, this new tool seems to have developed a specific competence to manage the text in the images, surpassing the capacities of some of its competitors such as stable Diffusion XL and Deepfloyd. In particular, Openai highlights her ability to produce convincing images of letters, figures, and even human hands (a real difficulty for these tools).

Dall-e 3 // Source: OPENAI

Dall-e 3 // Source: OPENAI

Dall-e 3 // Source: OPENAI

Dall-e 3 // Source: OPENAI

Dall-e 3 // Source: OPENAI

Dall-e 3 // Source: OPENAI

Dall-e 3 // Source: OPENAI

Dall-e 3 // Source: OPENAI

Dall-e 3 // Source: OPENAI

Dall-e 3 // Source: OPENAI

Dall-e 3 // Source: OPENAI

Dall-e 3 // Source: OPENAI

Dall-e 3 // Source: OPENAI

Dall-e 3 // Source: OPENAI

Dall-e 3 // Source: OPENAI

Dall-e 3 // Source: OPENAI

By integrating Dall-E at Chatgpt, OpenAi strengthens its position as a leader in the generative AI world. Chatgpt and Dall-E 3, together, promise to repel the borders of AI.

Chatbots will not only be able to converse in an almost human way, but also to imagine and create images from simple descriptions. We can’t wait to test.

If you want to test Dall-E 3, it will be available for Chatgpt Plus and Enterprise users. Which is much easier to use than Midjourney … where you need a Discord customer.

Chatgpt

To whom the images generated ?

The question of copyright linked to the images generated by AI is currently a hot topic. Current regulations suggest that IA images are entering the public domain. However, Openai clearly declares that the property of the images produced by Dall-E 3 returns to their creators. And that’s another debate in the sector.

Bing cat is also entitled to Dall-E 3

This Thursday, September 21, Microsoft held a conference to announce the very next arrival of Copilot, but also the new surface range. In parallel, the company communicated on the arrival of the Dall-E 3 model in Bing Image Creator. As a reminder, this is the Bing Chat images creation tool. According to Microsoft, Bing Image Creator ” Also understands what you are asking for, which allows you to get more precise images. »»

Microsoft also says go further by adding ” A digital watermark invisible to all the images generated by AI in Bing, including the time and date of their creation. “Difficult to know how powerful this watermark will be, nor how we can check its presence or not, but this is a first step.

To follow us, we invite you to download our Android and iOS application. You can read our articles, files, and watch our latest YouTube videos.