If you’ve used image generation within ChatGPT before, it’s probably been hit or miss. It still is, but it’s getting better.
OpenAI has used a tool called DALL-E for its image generation for the past few years. Recently, they came out with a new image generation tool within ChatGPT called 4o Image Generation.
Here are some of the improvements that we’ve noticed -
- Images are more consistent
- I.e., if you ask a follow-up question about changing an image, what you get will preserve more of the original image than before
- Text includes more real words
- A lot of the “text” before was nonsense characters. Now it actually forms words (sometimes)
- It is better at following instructions in the prompts
- It does a better job of creating hands
- Not as many 9-fingered people
It is still a work in progress, but they are moving in the right direction.
Since it is fairly new and still popular, it might take longer than usual to create your images, and they may cut you off sooner.
You don’t have to do anything special to use 4o Image Generation. When you ask ChatGPT to create an image, this is the default tool that it will call now.
Comments