Dall-E sounds like the name of some movie robot. And although it is somewhat related, in reality It is an artificial intelligence capable of generating entirely its own images.
This tool created by OpenAI has made a name for itself within applications for creating images using AI. But what exactly do you know about her? Here we explain everything you should know.
What is Dall-E
As we told you at the beginning of the article, Dall-E is a tool capable of creating images using artificial intelligence. It has been created for OpenAI and its operation is very simple since you only need to write what type of image you want and it will make several examples based on that description.
Its language model is GPT-3, very similar to ChatGPT, and has been trained with millions of parameters. That is why it is able to understand language and apply it to images. In addition, he has also received training with millions of photos and works of art, thereby He is able to draw famous personalities because he understands who they are and how to draw them.. Or he is able to use the technique of some artists to create original drawings.
The first version of Dall-E came out in 2021. However, it is not the one that can currently be used, but a year later they made modifications to improve the tool, and Dall-E2 came onto the market. Of course, a third vision is expected in a matter of a short time due to the great advances that this artificial intelligence is making.
How Dall-E works
As we have told you before, Dall-E is a tool capable of taking an idea that you explain to it and turning it into an image.
Once you write the text, as detailed as possible, and how you want the image to be, the AI encodes and interprets it to know exactly what you ask of it and see the characteristics that you have told it, where they should be, the style...
From that information already processed Dall-E begins creating the image information. It is not yet the image, but rather it begins to give instructions with a language that the AI understands to move on to the last phase, which would be a decoder. This is what causes the image to be painted through what has been asked of it.
In other words, When you ask for an image, first think about what you are asking to understand all the elements it must have.. And only when he understands the message you have left him does he get to work.
One of the doubts it generates is why when you ask for an image and it gives some results, if you ask it again for the same thing it generates other results. This is because it processes it from scratch. That is, it does not take into account what he has done before. That's why The first results can differ so much from the second and subsequent ones.
Dall-E is it free?
Given that Many artificial intelligence applications are starting to be paid, this question is not unreasonable.
But the truth is that we are still lucky to use it for free, at least for a while. What we don't know is for how long.
Of course, to use it you will have to go to the official Dall-E2 website and register. If you are already registered with ChatGPT, you can use the same keys to do with the images tool.
Once inside, you will find that you have some free credits to try it out and ask it to draw whatever you want.
In fact, you will have a text box to write the description of the image. Make it as detailed as possible and, if possible, better in English than in Spanish. We do not mean by this that he does not understand Spanish; but since it is a tool trained in English, you will have better results using that language.
When you click the Generate button you will have four images related to what you have asked for. Of course, if you like one but want to modify it, know that Dall-E will start from scratch, so you may not get the same result (even if you tell it that you want details of the results it has shown you).
Unless you download the image you liked and upload it to the tool to make modifications based on that image (for example adding elements, removing or replacing others...). Yes indeed, sometimes the result is worse than the one you first got For the same reason as before: it processes everything from scratch.
Once the credits run out you can recharge your account with new credits but to do so you will have to buy them. At first you will have 50 credits. And every time you ask for an image (even if it presents you with four) a credit will be consumed. Plus, you have fifteen free credits a month.
As for the price of the credits, it is not too expensive. Approximately it will be about 15 dollars for 115 credits.
This is what you can't do with Dall-E
It is clear that Dall-E is a quite powerful tool and well trained to make images of all kinds. But you may not know that it has a limitation. And the creation of images that could be potentially harmful is prohibited.
What do we mean by this?
You'll see, Dall-E will never take pictures of nudes or celebrities. Regarding the latter, you can use them as models, but you will not be able to ask them to take images of these famous people.
Nor will you show sexual or violent content in your images. Although we know that fooling the AI right now is quite easy, but it is something that its creator is working on to prevent the limitations they have imposed from being bypassed.
Are Dall-E results good?
Answering this question is not easy. If you ask for something simple, then yes, Dall-E can come very close to the image you have in mind. But the truth is that, if we are talking about something more complex, or something that came from your own imagination, you have to be very careful so that it comes close to what you have imagined. And that can mean wasting a lot of opportunities.
As you see, Dall-E can be a great tool for creating images. Have you ever tried it? What do you think?