Investigate (local) StableDiffusion support

[inFocus7]

We should support users who want to run as much as they can locally. There are many reasons why someone would want to do this:

1. Save $$$ by not using APIs. From my tests, generating 12 prompts + images costs ~$1.5 USD.
2. Generate content offline.
3. Not hit any content filters denying generations.

This is an initial investigation to create multiple issues as it would involve multiple steps, such as

• Investigating different local image GAN libraries (ex. stablediffusion a1111),
• Updating the UI/UX to cleanly support choosing between using DALL-E vs StableDiffusion
  • This portion could get messy since the prompting would also differ and we'd likely have a separate UI for each.
• (Optionally) doing the above + refactoring to allow plug-ability of more support in the future.

https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/API

• in a nutshell, could probably curl to its localhost port.

• doing a GET: /sdapi/v1/sd-models returns available models, which i could display a name in the dropdown mapped to the full model name (theres a name and title field in the return)

• could then, based on user input, perform a POST: /sdapi/v1/txt2img with all prompt data + chosen model and size (might be opinianated on sizes to standard cover 512x512 sizes and portrait).

• note since a1111 ui + this are gradio they both get exposed on the same port, so will need to change mine.