Openai recently announced an important novelty for Chatgpt: now the chatbot based on artificial intelligence can generate images in a native way based on the GPT-4O model. This innovation represents a significant step forward compared to the previous GPT-4O skills, which until now were limited to the generation and modification of textual content. GPT-4O technology promises more accurate and detailed images than its predecessor, From 3even if the generation process takes a little more time. The generation of images with GPT-4O is also available on Sora. The system is able to modify existing images, including those with people, intervening on elements in the foreground and in the background, all surprisingly accurately.
What can the integrated GPT-4o generator of images do: the news
The novelty announced in the past few hours by Openai represents an important change for the generative artificial intelligence sector. Until now, in fact, Chatgpt had distinguished itself as an advanced chatbot, but its visual processing capacity was delegated to separate tools as by the-e. With GPT-4O, on the other hand, The creation of images has become an integrated and more sophisticated process. The model was designed to guarantee a higher level of detail and visual loyalty, improving the yield of the texts in the images – unlike what happened with the previous models, where writing was often distorted or illegible – and opening new perspectives in automated photo editing and in the creation of personalized visual content thanks to the ability of the model to modify existing images.
Another noteworthy technological aspect concerns the method with which GPT-4o generates images. THE’“Authorized” approach employed by the model allows you to produce images with a sequential logicworking progressively on the different sections of the figure, instead of generating the entire content in a single step. This procedure, although slower than models operating in parallel, contributes to improving the quality and precision of the final result.
If on the one hand this innovation opens up new creative possibilities and simplifies the production of visual content, on the other raises important issues on the management of copyright, on the accessibility of these tools to all users and also on the difficulty of distinguishing real content from artificially produced images.
Regarding the issue relating to copyright, to train the model, Openii used public data as well as images provided by partners like ShuttersTock. The company has assured that she respects the rights of artists, preventing the generation of works that directly imitate the work of living creators and offering an opt-out module for those who want to exclude their work from the training datasets. Despite this, the training of the model on public data and images from commercial partnerships represents a controversial and delicate aspect. Openii, like other companies in the sector, maintains the specific details of the dataset, both to protect their competitive advantage and to avoid legal disputes related to copyright. And all this raises doubts about the real respect for the copyright of photographers, illustrators and graphic designers from all over the world.
Copyright question aside, if you want to see the Openai images of GPT-4O images, we leave you to the following introductory video made by the company.
https://www.youtube.com/watch?v=e9n8jx-uc
Delayed availability for free users
Currently, the generation of images using the GPT-4O model is Available for Chatgpt Pro Plus, Pro and Team subscriberswhile the release for free users will undergo a delay due to the high question, as confirmed the same Sam AltmanCEO of Openai, who reported:
The chatgpt images are much more popular than we expected (and we had quite high expectations). The launch of our free level unfortunately will be postponed for a while.
