In just two years, the artificial intelligence sector is unrecognizable; It has gone from being a futuristic technology whose utility was in doubt, to a mass product that all the big companies bet. And a good part of this turn has OpenAi, with Chatgpt, which was the ‘chatbot’ that demonstrated the potential of the generative AI.
Since then, things have changed a lot. The number of competitors in the sector is very high, and every day is greater; And this increase in competition assumes that what was previously incredible is no longer enough. Chatgpt has suffered especially for this, with rivals such as Gemini and Deepseek offering more functions and similar performance at a low price.
OpenAi’s reaction has not been expected, with important internal changes that will be noticed over the coming months, and that means making shocking decisions such as abandoning new models to focus on futures. Today, Openai has announced another of these changes, which simplifies its line of product and potency chatgpt: the Creation of images by AI.
Of course, Openai already had a model of image generation, and one of the most popular: we talked about Dall-E, which was one of the first generators of Ia-based images that really demonstrated the potential of this technology (for better and for worse). However, now Openai has decided to focus on chatgpt, which now wins the functions of generating images with the New 4th model by default, abandoning Dall-e 3.
With model 4, Chatgpt is capable of generating images natively And without the need for an external model, in addition to allowing the edition of advanced images. The new model also brings important improvements with respect to Dall-E, and specifically, solve some of its biggest problems such as text generation.
Image generated by chatgpt
The free android
A very common failure in the first images generated by AI was in the texts; The models were simply not able to understand the way they had to insert the characters into the image. Therefore, the great advance of 4 is precisely In the texts that appear in the imagewhich are now rendered precisely and following the user’s instructions to the letter, according to OpenAi.
The free android
In this way, it is easier to create the image we really want, without creating a basic image and then putting the text using an image editor. The result should be more realistic imagesS, in which the text is better integrated with the rest of the stage, and to demonstrate it, OpenAi has published some surprising examples.
The new 4th model also brings important Improvements in the consistency of the characterswhich is important if we are going to create several consecutive images. For example, if we ask for an image of a woman writing, then we can ask him to make an image of that same person and his appearance will be the same, instead of creating another person. In the same way, the generation of shift images has also been improved, in which the user refines the image by adding or removing elements in each step.
Chatgpt’s new 4th model is now available for all users, including those who have free accounts, in addition to Plus, Pro and Team payment accounts. Access to Dall-E is maintained, with a new Dall-E GPT. In addition, the new model can also be used in Sora, AI to generate videos.
