ChatGPT’s new image generator is currently among the best out there with the ability to not only create pictures from scratch but also to turn real life images in various creative styles. While ChatGPT cannot convert the images in the style of living artists, it can turn the real life images of users in the style of various studios like Pixar, Studio Ghibli and even TV shows like The Simpsons.
While Ghibli-style AI images dominated the internet for a while, the trend has now gradually shifted towards generating personalised AI action figures—reminiscent of the superhero-style toys often found in toy shops.
We have already covered how to create personalised action figures using ChatGPT. Now, with the ChatGPT’s improved storage capabilities, the ChatGPT offers an even more personalised experience.
What is ChatGPT’s memory update?
ChatGPT can now remember the previous conversations it has had with users to customize its responses, taking into account the user’s preferences. The chatbot uses its ‘memory’ to adapt its tone and respond to the interests or recurring topics mentioned by the user.
While the new update may not seem like much on paper, my experience with ChatGPT over the past few days suggests that this is perhaps one of the most powerful updates the chatbot has ever received. Gone are the days of getting a generic answer to all your questions. Instead, ChatGPT now knows what I do for a living, what my hobbies are, and even my personality type (scary).
These new capabilities can now be used alongside GPT-4o’s native image generator to bring action figures to life using original user images.
How to generate personalized action figures using ChatGPT?
1) Open ChatGPT app or website and make sure the model is selected as GPT-4o
2) Upload your desired image by clicking on the “+” icon
3) Enter this prompt in the text box: “Create an image of me as a real-life action figure. Be specific as possible based on what you know about me. And put multiple interesting things that represent who I am inside the packaging. I’ve attached an image of me for reference.”
4) Do not worry if you don’t get the picture perfect image that you had in mind in the first attempt. Instead, you can tell ChatGPT to make the specific changes or even ask the chatbot what mistakes it thinks were made in the image. When you analyze all the mistakes, ask ChatGPT to make the required changes.
Again, it might not seem like a major thing, asking ChatGPT to reference the past conversations but the personalized action figure images look very different from the ones that are generated in generic style. The first action figure below is one which is personalized with ChatGPT while the other two are generated by giving detailed prompts in specific style.
What ChatGPT gets wrong?
ChatGPT image generator does tend to get the facial features horribly wrong with absolutely no resemblance to the original image. This is something I have noticed with various filters applied using GPT-4o’s image generator and is not an isolated issue with many social media users facing similar concerns.
Another thing ChatGPT gets wrong is that some times a text or an object can be repeated. Other times, the text is simply not clear or makes no sense (Compare the first image to the one given below).
The ongoing problem suggests that while ChatGPT has come a long way from relying on DALL-E 3 for image generation, it still has a lot of room for improvement.
