ChatGPT has updated its image feature, and it is now powered by the company’s new flagship image generation model. Not only is it faster than before, but it also brings added convenience and is now more precise than ever when it comes to edits. You can describe things in very fine detail. If you want to change something about an image, you can pinpoint that you want just one specific thing changed. I have been using the feature for the past few days, and I finally have some thoughts to share.
I think it is much faster than before, and there are some major conveniences at play here as well. So, if you are clueless about where to start and just want to experiment with the feature, there is the new “Try on a style on an image” section, the Discover section, and a convenient place to access your images. I have had a fun time using the new images upgrade, and I think you should definitely try it. Here is what I thought after using it for a few days.
ChatGPT Images is now faster and easier to work with
One upgrade I immediately noticed with ChatGPT images is that it is easier to get ChatGPT to follow your specific prompt. So, if you clearly state that you want it to use bold, saturated colours for the font for, let’s say, a Bollywood-style poster, it will actually stick to that and will not stray from your original prompt. There is less hallucination, and it also maintains your tone quite well. So, if you tell it to stick to a retro design, it will actually follow that, which is a really good thing.
Another thing I especially liked about the new images feature is the Try on a style option. This means you do not have to craft a prompt out of thin air and expect incredible results, because in most cases, when you end up with a bad-looking AI-generated image, the prompt is usually to blame. Here, OpenAI provides prompts of its own. You can try various styles such as 3D Glam Doll, Doodle, Fish Eye, Ink Work, Ornament, Pop Art, Sugar Cookie, Bollywood Poster, Jaipur Textile, and more.
You also have the Discover tab, which gives you ideas to create things like matching outfits for couples, professional product photos, restoring an old photo, and more. I tried all of these, and they were thoroughly fun and useful. So, if you actually need them, you know where to go.
What is the quality like?
Right off the bat, I can tell you that things like combining images are much better and more accurate. I tried combining my photo with a Bollywood-style poster, and ChatGPT did a great job. It looked very lifelike. Another thing I immediately noticed was that once ChatGPT is done creating an image, you can follow up with further changes by giving it another direction. If you are not happy with something in your poster and want to change it, for example, if you want to add more destruction to a Bollywood-style poster, you can simply ask ChatGPT to do so. You can also ask it to add more objects to your image by submitting a new photo and asking ChatGPT to include that object in the original image.
Another improvement I noticed is that text rendering is much better. Google Gemini Nano Banana’s latest update has also improved its text rendering capabilities, and ChatGPT images has now improved in this area as well.
What could have been better?
One thing I immediately noticed with this update is the facial rendition. While it is better than before, it is still not quite there, especially when you want to place yourself in a different scene. If you want to put yourself into, say, a dramatic movie-style scene, it recreates your face and facial details fairly well, but it is not a one-to-one match, at least based on my own experience. In some instances, it does a very convincing job, while in others, it does not. That is one area where I think ChatGPT images can certainly improve.
However, if you are transforming yourself into a completely different art style, such as comic-style or pop-art, the model does a fairly good job and is a big improvement compared to before.
MOBILE FINDER: iPhone 17 LATEST Price
