AI for creating visual worlds with text is dramatically more powerful with ChatGPT

How writers can use ChatGPT to generate ever more wonderful images with AI

The speed with which generative AI is improving is so fast that making any predictions seems perilous. However one astonishing power of the newly released GPTChat is its ability to suggest prompts for generative image software like Picto (based on stable diffusion) or Midjourney or Dall-E

In the context of what we are building on stck.me, this is a powerful way for writers to convert their ideas into visuals. And in the long run entire new worlds. For example here was my ask of GPT

Using ChatGPT to generate image prompts is powerful and fun

And here are some of those prompts converted into images (with no edits).

A flower with a glowing heart
A fantastical flower with rainbow-colored petals that swirl and twirl in a never ending swirl

Another more ambitious ask from chatGPT, “Suggest a few prompts to create birds more beautiful than any found in nature. Use real birds as starting point to build a variety of new magnificent species.” The bird in the cover image and one below is generated by text to image AI.

ChatGPT tries to imagine new kinds of birds (yes it’s limited to what it knows)
A majestic bird with iridescent blue and green feathers and a crown of golden plumage on its head. Its eyes sparkle like gems, and it has a powerful beak that can crack open even the toughest nuts.

This is Year One of widely available generative AI. And from here it is not hard to imagine a version where entire storylines can be converted into art, comics, animation, or even realistic videos. The line from here to that world is no longer a conceptual challenge but a matter of time and cost.

The debate about whether the AI can imagine something entirely new seems moot. For now, the fact it can help us convert words and ideas into images is magical enough.

Write a comment ...

Write a comment ...