If you are familiar with generating images with AI models, it's possible you already know that these models sometimes lack the ability to create images that reflect specific directions given in your prompt. Well, this might be changing as Open AI has announced a significant upgrade to ChatGPT's image-generating capabilities, which will help it follow prompt directions with increased accuracy. For reference, the above image was created by ChatGPT with the following prompt, "draw me a futuristic image illustrating your new image generation capabilities." Not too shabby, we'd say.
A few hours ago, OpenAI's official Twitter account announced that it had launched its new image model - 4o image generator, which will be available for subscription levels including "Plus, Pro, Team, and Free users.” However, this feature has not been rolled out for users on the education and advanced security subscription levels - Enterprise and Edu.
For AI image generator users who have noticed that most of these models struggle to render text accurately, OpenAI has promised that mistakes such as inaccurate capitalization and spelling errors common with AI image generators will likely not exist in this model. This will go a long way for some users because many have noted that, just like most AI image generators, OpenAI's DALL-E 3 does not always accurately reflect the details in the prompts.
To showcase the ability of the newly released 4o image generator, OpenAI released a series of demo videos on an X thread where a representative of OpenAI demonstrated that adding the 4o image generator will help ChatGPT capture a nearly perfect representation of imputed prompts. In one of the videos, a representative tasked the model with generating 15 images with different attributes. The model generated outputs that suggested the AI had paid close attention to most of the details in the prompt.
Image Credit: OpenAI
In another demo video, a representative demonstrated how users can customize images to address specific needs. A representative used the model to generate a transparent picture of a cartoonize puppy, and the output was relatively impressive. OpenAI also explained that you simply need to “describe what you need, including any specifics like aspect ratio, exact colors using hex codes, or a transparent background.”
While we're not sure that adding a 4o image generator to ChatGPT will help it accurately do something as critical as. say, creating an ideal logo for your dream business, its improved attention to detail will probably encourage more users to deploy the model's power to create and remain productive.
