dreamwalker

prior to the integration of image and edit models directly into chatbots, I created a tool that allowed you to explore the output of diffusion models in a context that de-emphasized prompting and instead forwarded the idea of exploration.

users could generate an image, and then request a “change” to the image in the form of a simple prompt or phrase, that would then transform the underlying prompt with that request and produce a new prompt that was then fed into the generator.

these images would then form a graph of connections with each other, allowing you to revisit older ideas or push forward new ones but maintain a growing web of generations around a given idea.

I may remake this tool again to work on the web… maybe…

docs