During on Tuesday live, the general director of Opeli, Altman himself announced the first large update of the possibility of generating ChatGPT image for over a year.
CHATGPT can now use the company GPT-4O model to natively create and modify images and photos. GPT-4O for a long time at the root of the AI chatbot platform, but until now the model has been able to generate and edit only text-not images.
Altman said that the native generation of GPT-4O image is live in ChatgPT and Sora today, the AI OpenAI video generation product, for subscribers of the PRO plan worth USD 200 per month. Opeli claims that the function soon appears on the plus and free chatgpt users, as well as programmers using the company API.
GPT-4O with the output of the image “Thoughts” a little longer than the image generation model, which it effectively replaces, Dall-E 3, so that what OpenAi describes as more accurate and detailed images. GPT-4O can edit existing images, including images with people in them-educating them or “internal” details, such as objects in the foreground and background objects.
To power the new image function, Opennai said Wall Street Journal He trained GPT-4O in the scope of “publicly available data”, as well as reserved data from its partnerships with companies such as Shutterstock.
Many generative suppliers of artificial intelligence perceive training data as a competitive advantage, so they retain it and all information related to it close to the chest. But detailed information on training is also a potential source of IP -related processes, another discouraging for companies.
“We respect the rights of artists in the scope of how we perform the results and we have rules that prevent us from generating paintings that directly imitate the work of all living artists,” said Brad Lightcap, Openai Operational Director, in a statement for the magazine.
Opeli offers a resignation form that allows creators to ask to remove their work from training sets. The company also claims that it respects requests to reject its bots from collecting training data, including paintings, from websites.
The improved function of the CHATGPT image generation occurs on the heels of the experimental native image of the Google image for Gemini 2.0 Flash, one of the company's flagship models. The powerful function has become popular in social media – but not necessarily for the best reasons. It turned out that the Gemini 2.0 Flash image component has several handrails, allowing people to remove water signs and create images depicting figures protected by copyright.
This article was updated at 12.00 PT to contain OPENAI statement to the Wall Street Journal around GPT-4O training data.
