#chatgpt
Thread
Playing around with ChatGPT's new image generation. Prompt: "create images for a 5 slide deck on the topic of scaling a customer community". Um... cool graphic, but the text looks like it was prompted with "write this in the style of a drunk college student returning from a party". :woman-facepalming:
DALL-E is probably the best image generation tool at text so far, but it's definitely more focused on imagery than textual content/meaning.
The way you might come at this to get something less yikes:
First, ideate:
"I am creating a presentation on X. Here's the content for slides X, Y, Z. What types of imagery might make sense to complement the points I'm making on those slides?"
Second, create:
"Now, I want to create imagery for those slides, one by one. I'm going to give you the slide number I referenced above. I want all of these images to be consistent in style: (describe the style, color pallet, etc.).
Let's begin with slide X"
I wouldn't try to combine the text and the image -- there's just no real control over fonts, and while it can get things right pretty often with things like headlines/taglines, immediately breaks on anything text heavy.
For example, it did a good job with this simple image for a LinkedIn post!