Using ChatGPT to create prompts for Stable Diffusion
ChatGPT has brought the power of language models for dialogue to the masses. You’ll find lots of use cases, and also its flaws, all over the internet.
But can you use ChatGPT to create unique prompts for Stable Diffusion?
The quick answer is “yes”
It all starts with what you ask from ChatGPT. The image above was generated from the prompt ”Time-traveling robot gardens”.
But how did ChatGPT came up with such a strange prompt? Simply by asking it to “Create surreal phrases with 3 words”
ChatGPT returned the following:
- “Electric jellyfish circus”
- “Galactic unicorn manifestation”
- “Hypnotic mushroom playground”
- “Neon dragon dreamscape”
- “Psychic octopus meditation”
- “Time-traveling robot gardens”
- “Flaming phoenix constellation”
- “Invisible elephant orchestra”
- “Rainbow-colored dinosaur parade”
- “Serene mermaid waterfall”
Depending on the model you are using, the results will differ for these prompts, of course.
The results above were obtained by using Diffusion Bee with the dream like art model
Conclusions
As explained you can use ChatGPT to come up with simple, or complex prompts, and then feed them into Stable Diffusion and get amazing results. Not all of them will be amazing, of course. In the prompts I used we go from the really detailed to results that fail to capture the prompt. One more reason to keep exploring.
