I wonder what would have happened if stablediffusion released itself as a really powerful web-based image editing tool, still offering but relatively de-emphasizing its "create things with a prompt from scratch" functionality.

Realistically, practical use cases are going to be AI+human mix for quite a while anyway.

We are building an advanced editing tools built on stable diffusion at artifactai.co, targeting creative professionals. Interestingly, we are seeing much more demand for generating image from scratch than edits. Emerged use cases: 1) quickly go from idea to visual presentation for brainstorming solo or w/ team, and…

and 2) generate variations of their own work with no marginal cost by fine tuning models using their styles and characters.

In both cases, it’s human + AI to your point even tho generating images from scratch.