Chat GPT has recently introduced a “Create Image” feature, and it’s powerful.
By selecting the “Create Image” mode in Chat GPT, you’re basically getting it into “visual-blender” mode.
Chat GPT will review your text instructions + any images you’ve shown it, and then Chat GPT will produce a new image based on everything you told it.
To start having fun with the ‘Create Image’ feature, you’ll need to do two things:
– First type some text to Chat GPT that describes what you want to achieve.
– Next upload some images to Chat GPT which will serve as an image-reference that GPT uses to build your new image output.
For a real-world example, the specific ai meme below was created using Chat GPT, using the method described above.
The starting point for the cartoon depicted above was actually based on a dry-erase board doodle.
I showed the doodle to Chat GPT, and gave it some instructions for how I want my the doodle to be re-illustrated.
Chat GPT’s first image result was… fine.
Chat GPT’s next image was better, but not perfect yet.
There were two small-but-important things I wanted to change, so I used Adobe Photoshop‘s ‘Generative Fill‘ feature:
1. I oriented the audience so that they’re going into the building, instead of away from it.
2. The robot is holding a violin, in a dejected fashion.
I think there will eventually be a point when Ai actually does have a sense of humor, and I think that will be a very new kind of Ai.
For now, I’m having fun making images with the Create Image feature of Chat GPT.
The ‘Create Image’ features I’ve described are available to all users, free and premium accounts alike.
Have fun!
Here are a few other examples I’ve made, using this ‘Create Image’ feature in Chat GPT:
The image above was created by giving ChatGPT two screenshots from the movie Timecop, and some text to guide our desired style of the resulting image:
Be specific about what you want to see in your image, and which pieces should go where; use “quotes” around any text that you to see in your final image output.
For this next image, I tasked ChatGPT with creating an image that depicts “how to bake a pizza“, based on real photos I had taken of a real pizza I made.
I uploaded 9 images to ChatGPT in order to create that image of the pizzas above.
As you can see, the image and it accompanying text all appear correctly, although there is no step that says “add cheese”, but we can consider cheese under the ‘add toppings’ headline.
The text I typed to Chat GPT was clear and specific about ‘how’ I wanted the output image to look.
When it came to the specifics of the image output (such as the text, order of the images, etc) I just let ChatGPT figure things out.
