First I had this conversation with NovelAI; its contributions are in bold:
“So I want to have one AI produce a paragraph, and then feed that to another AI to produce an image from.”
“That sounds interesting,” the AI console typed back, “would you like me to generate a paragraph of interesting text, to get you started?”
“Sure,” I said, “can you do that”
“Easily,” the AI replied, and then it emitted this paragraph:
A man walked down a road. He was very old, but he had been walking for many years now, so his age did not matter much any more. The sun rose in front of him as he walked along, its light warming his face and making the wrinkles on his skin seem less harsh than they might otherwise be. His eyes were still clear and bright with intelligence, even though those around him often looked at them without seeing what lay behind them. He saw things other people missed; he knew how the world worked and why it worked that way.
“Wow, thank you!” I thanked the AI, “I’ll have to break that up into multiple prompts, but I can’t wait to see what VQGAN+CLIP comes up with from it!”
The next day, I fed the generated text to my CLIP-based AI, which produced these images:
I thought that last line was amusingly appropriate, although of course there was no reason to wait until “the next day”. I fed the generated text immediately to Nightcafe, and it produced these images (same prompt text for each, just different random seeds):
I kind of love it! See the face in the clouds in that last one?
ceoln 