See for Yourself: ChatGPT's New Image Generation Is Insanely Good
1 min read
Summary
OpenAI has released a new image generation model which is integrated into ChatGPT 4o
The new model is far superior to DALL-E in being able to correctly identify images, particularly around hands and faces
-Licensed images aren’t allowed but images similar to licensed characters are possible with the new model
asks a user to compare images it has generated with those generated by the original DALL-E model, and in many cases, the new model is far superior.
Some correct images generated by ChatGPT 4o include a guitar chord, Albert Einstein eating ice cream, a Sith Lord hailing a taxi, and Geralt of Rivia shopping for groceries.
It also does well with cartoon-style illustrations, mirrors and reflections, and text and letters.
However, it falls down on images involving cars and streets, where a Peugeot 206 is depicted incorrectly parked on the sidewalk.
