The Dall-E image generator is now available for the general public-digital, discovers the secrets of Dall-E

The secrets of artificial intelligence Dall-e 2

Above all, you must keep in mind that, whatever the power of artificial intelligence, it will not be able to replace human creativity. That being said, Dall-E 2 offers an impressive amount of possibilities, which may suggest that AI can be added a real place in the world art, in the near future. Dall-E 2 is what is called an “artificial intelligence algorithm”. It is based on a network of artificial neurons which allows you to generate images from short, but precise sentences. It is a practical tool not only for content creators, but also for all people with an artistic spirit who want to try something new. It can also be used for university work or artistic projects, for example.

The Dall-E image generator is now available for the general public

Very popular in recent months, Dall-E has just opened to the general public. Anyone can now ask artificial intelligence to produce wacky, disturbing or amazing graphic creations.

Advertising, your content continues below

Dall-e, the famous artificial intelligence capable of transforming textual suggestions into photorealistic images, is now available for the Quidam. Reserved so far for a handful of handpicked users, the tool published by Openai is accessible to all.

A simple interface

No need to register on the waiting list and receive a formal invitation: the site now offers anyone who wants to provide an email address to generate images via artificial intelligence. The interface could not be simpler with a field of text in which we formulate suggestions. After a few seconds of waiting, Dall-E presents four images related to the suggestion. It is then possible to modify them or create variations related to the same theme.

The tool is impressive, generating photorealistic images or imitations of art work in a few clicks. Obviously, Dall-e is doing particularly well when it is subject to specific requests. “A painting of the movement of yellow vests in the style of Eugène Delacroix” Allows for example to see how much the machine has assimilated the style of the painter. By subjecting a more vague request, such as “Digital”, The result is much more baroque and strange.

Advertising, your content continues below

Well aware of the risks presented by its AI in the dissemination of infox and dubious assemblies, Openai has installed fairly strict safeguards. Impossible for example to offer suggestions containing “Emmanuel Macron” or other known public figures. Nudity and violence are also banned, and any research containing references to these concepts will result in a page indicating that the suggestion is contrary to the rules of the platform. “In the past few months, we have strengthened our filters in order to block attempts to produce sexual, violent and others that violate our content policy”, Indicates the company.

The problem of algorithmic biases

“The machine must be confronted with reality to learn”, explains Sam Altman, the CEO of Openai, to justify the opening of Dall-E to the general public. The subject is particularly sensitive, because the latter has previously been accused of strengthening racist or sexist stereotypes when it comes to representing a person with power and responsibilities. To avoid producing only images with white CEOs in fifteen years, it would seem that Dall-E adds invisible keywords (black man, Asian woman) in order to diversify the results.

This wobbly method to say the least, aims to counter the biases inherent in any artificial intelligence system. Dall-E was fed with a lot of right-free images available on the net, and the latter often tend to present a somewhat stereotypical image of the company. More simply, Dall-e has sexist and racist biases, because the world in which it was built has sexist and racist biases.

Any registration for Dall-E comes with 50 free credits (1 credit = 1 image generation) and 15 credits are added to your account each month. A large amount of sufficient to start testing the machine. It is also possible to buy credits by package of 115 for $ 15.

Advertising, your content continues below

Follow all digital news on Google News

The secrets of artificial intelligence Dall-e 2

By filling out this form, I accept to receive the newsletter from Etudestech and I understand that I can unsubscribe easily at any time at any time.

Dall-E 2

You know chatgpt by Open IA ? Did you know that the company created another artificial intelligence text-to-art which allows you to transform a text into an image ? These are Dall-E 2. Recently, it has been accessible to everyone and studies tech reveals everything you need to know about it: the principle of this artificial intelligence (AI), how to use it, what are its limits and above all, what is its price.

What is dall-e 2 ?

Initially, Dall-E 2 was created in April 2022, with a beta version accessible to a small group of people. If you wonder where this name comes from which reminds you of something, the answer is simple. The name Dall-E is a suitcase word between the painter Salvador Dali, and the animated film Wall-E, by Pixar. After his first test, Open IA launched a second beta version, this time, accessible to all by registering on the waiting list. It is often said that an image is worth a thousand words, but Dall-e 2 only requires a few ! Simple, fast and practical, the tool allows you to create a high quality professional image thanks to a written description, which follows some rules. Concretely, the tool is based on OPENAI GPT-3 AI, capable of understanding and treating human natural language to convert it into an image.

Above all, you must keep in mind that, whatever the power of artificial intelligence, it will not be able to replace human creativity. That being said, Dall-E 2 offers an impressive amount of possibilities, which may suggest that AI can be added a real place in the world art, in the near future. Dall-E 2 is what is called an “artificial intelligence algorithm”. It is based on a network of artificial neurons which allows you to generate images from short, but precise sentences. It is a practical tool not only for content creators, but also for all people with an artistic spirit who want to try something new. It can also be used for university work or artistic projects, for example.

The promise of Dall-E 2 is simple: creating art generated by AI with a minimum of effort or human intervention thanks to 650 million possibilities. For the moment, this promise is held, even if obtaining a satisfactory result requires control of certain techniques.

How to use Dall-E 2 ?

You will have understood, Dalle-e 2 aims to allow everyone to access the text generator in image, with ease. The use of the Dall-E 2 program is approximately the same level as a search for images on Google. You just have to type a short sentence in the tool search bar to generate an image. The big difference lies in the fact that Dall-E 2 creates all new images using a combination of artificial intelligence (AI) and automatic learning (machine learning). Here are the steps to start using Dall-E 2 and wake up the sleepy artist in you.

First of all, you must start by going to the official Dall-E 2 website, then click on “sign up” allowing you to create an account. To save time, do not hesitate to register with your Google or Microsoft account. Otherwise, you can just do it with your email address and a password. In all cases, the registration process requires two factors authentication: by SMS and by email. Once your identity has been confirmed, you can click on the “Continue” button to accept the conditions of use.

Once it is all over, a welcome message is displayed and Open Ai quickly explains what you can do with the artificial intelligence program. After having read the many possibilities available to you, pressing the “Continue” button.

To create a new image by artificial intelligence, you just have to enter a brief description of what you have in mind, and what you imagine. Dall-E 2 allows you to exploit up to 400 characters (more than a tweet which has only 280 maximum) to describe what you want to see appear. Slab created four images per request with a different style for each. Then, you can click on each of the four images to make basic changes, such as adding borders or erase certain parts. Otherwise, you can also ask Dall-E 2 to create different versions of this image. It is important to keep in mind the simple fact of modifying a few words here and there can make all the difference, because everyone has a specific meaning. It can also be useful to explore different types of art, such as digital art, portraits, 3D rendering, but also sketches. Overall, the more you train and the more the result obtained will be likely to be credible and realistic.

There are many options with Dall-E: for example, you can click on the “Surprise Me” button (surprise me), or download an existing image and ask the program to modify it according to your preferences and your expectations. Functionality Outpainting, Recently added to Dall-E 2, also makes it possible to extend an image beyond its original borders. It can be applied to an image created by AI, or on an image that you have downloaded beforehand before. The tool has already been tested on famous works of art like the Mona Lisa, and has made impressive results.

To use this function, you just have to generate or download an image, then reduce your size by pulling the angles. Write your ” prompt “, And Dall-e will take care of adding the elements by taking up the style of the original work.

Improve your images on slab 2

If you have already tried to use Dall-E 2, you could see that the result obtained is not always up to your expectations. To improve your creations, you can use the ” Prompt book From Guy Parsons, a visual resource specially dedicated to art IA. It helps in particular in the formulation of textual descriptions, and can also inspire you to exploit all the potential of Dall-E. The objective is that the descriptions you formulate are easily understandable by artificial intelligence.

In the guide, you will find 82 pages which detail the best techniques to perfect your results on Dall-E. One of the most important advice relates to the choice of adjectives used to grasp the atmosphere, emotion or the aesthetic composition sought after. That’s not all, the prompt book deciphers tips for all types of images: photography, portraits, landscapes, but also different styles of illustrations and historical arts as well as 3D art, more complex to realize. The book also reveals advice for viewing angles, lighting, type of objective, or textures. This is a tool not to be overlooked if you want to improve !

How to delete the watermark ?

You have surely noticed it if you have already tried in artistic creation with Dall-E 2, the images generated by the AI ​​contain a signature resembling a line of colored squares located at the bottom right of the image, which makes them recognizable. If this could quickly annoy some users, this era is over. Indeed, the Dall-E 2 regulation now authorizes to remove this watermark, which is essential for uses for commercial purposes, for example. It is possible to remove the signature very easily with any image editing application, such as Photoshop. If you have no idea how to do it, know that it is also possible to download the image directly without watermark.

On a computer, you have to choose the “inspect” option and search for the Windows URL.net, right -clicking. Then copies the image link and open it up. The image should appear without the logo. On smartphone or tablet, you can directly press the image on the generation page, and click on “Save the image”.

Dall-E 2 limits

An understanding of words still limited

The main limit of Dall-E 2 is that the quality of the rendering depends largely on the text provided by the user. The more precise you will be, the more likely you will be to obtain the desired result. That being said, the system has certain limits. For example, Dall-E 2 does not often manage to merge several objects or properties of objects such as shape, orientation and color. In addition, if data labeling done on Dall-E 2 is incorrect, it can produce false results, like someone who would have learned the bad word. Concretely, artificial intelligence is being created, so it is possible that it does not know certain words or certain turns of sentences. These are problems that will disappear over time and the next updates.

You should know that Openai to take care to program Dall-E 2 so that no real face, nor real person, is taken up in image creation. The objective here is to prevent the abusive use of the program. That said, more and more similar tools exist, which makes it possible to take an image of Dall-E and integrate the face of someone else.

More Dall-E 2 gains in popularity, the more copyright question arises for users. Even if the company Open AI has attested that users ” benefit from all the marketing rights of the images they create with Dall-E, including the right to reprint them, sell them and market themR “, art generators by AI are based on existing data. Even if it is not intentional, the option of violation of the intellectual property right could, in the future, be problematic.

Dall · E 2: what is prohibited

The priority of Open AI for Dall-E 2 concerns above all the questions of ethics, in particular with the representation of diversity and respect for the limits imposed on other types of images, generated by humans. Strict rules have been implemented to avoid “bias and toxicity” of the images generated by Dall-E. The genre and the ethnicity are chosen randomly if they are not specified in the description. Dall-e will automatically reject images containing realistic human faces or resembling public figures such as stars or politicians, who could be integrated into the images generator.

Also, Openai does not allow the creation of images considered to be offensive. This is particularly the case with images illustrating self -harm, hateful symbols or illegal acts (weapons, crimes or drugs). Automated surveillance systems and human moderators are responsible for censor prohibited content.

Is Dall-E 2 ?

Dall-e 2 is unfortunately not completely free. Image artificial intelligence has a credits system to use it. When registering on the site, users receive 50 credits which therefore limit their use of Dall-E 2. Then each receives 15 credits per month, without paying. Each credit corresponds to a request made on Dall-E 2 and therefore to the 4 images obtained. If you have not consumed your 15 credits during the month, they will not be postponed in the next month and you will start again at 15 credits maximum. If you have no more free credits, but you would like to explore Dall-E 2, don’t panic. It is possible to buy 115 credits by paying the sum of $ 15. Unlike free credits, the credits purchased are valid 12 months from the date of purchase. Each generation of the new version of an image will cost you a credit.

Where to download Dall-E ?

Initially launched in April 2022, Dall-E 2 was only accessible on the waiting list for five months. Since September 2022, access is now open and anyone can register from the official website. So you don’t need to download anything to use this artificial intelligence. You just have to register on the site as explained a little higher, then let your creativity express themselves.

Open Ai, what’s what ?

It is no longer a surprise for anyone, Open AI is the company behind the creation of Dall-E 2, but also Chatgpt, a conversational artificial intelligence that appeared very in vogue. Openai was founded in December 2015. Originally, the idea was formulated by Elon Musk, Sam Altman, Greg Brockman, Ilya Sutskever, Wojciech Zaremba, as well as other influential people in the field of technology.

The organization was created to respond to a shared concern about the potential IA negative impact, and the need to guide its development in a responsible manner. In April 2016, Open Ai published its first research document called “A Association”. In this document, the organization underlines the importance of developing an artificial intelligence “beneficial for humanity” while avoiding potential risks. Subsequently, other research was published by the company to promote collaboration, but above all encourage transparency in the field of AI.

In 2018, Openai announced the launch of its flagship project: GPT (Generative Pre-Trained Transform). It is a model of language model based on deep learning. The second GPT-2 version is, on the other hand, unveiled in 2019 and attracts very much attention. And for good reason: it surprises by its ability to generate text in a creative and coherent way.

In parallel, Openai continues to do research and takes measures to advance the security of the AI, in particular by participating in security competitions, by encouraging research on robustness and promoting the adoption of good practices in the domain. The organization also emphasizes collaboration with other actors in the AI ​​field to collectively advance research and ensure ethical and responsible standards in the development of AI.

In 2020, Openai launched GPT-3, the version that will arouse the greatest craze around these technologies. Since then, Chatgpt-4 has emerged and does much more: creative and technical writing, song composition, writing scenarios, etc. Open AI explained in a press release: ” We spent 6 months to make GPT-4 safer and more aligned. GPT-4 is 82% less likely to respond to requests for unauthorized content and 40% more likely to produce factual responses than GPT-3.5 During our internal assessments. »»

Thanks! You've already liked this
No comments