AIProxy now supports Flux-Dev ControlNet by xlabs-ai. You can use this model for image-to-image generation. In our experience, inference takes around 3 minutes if the replicate model is cold and 30-40 seconds if warm.

There are loads of controls available to tweak your image, so please take a look at `ReplicateFluxDevControlNetInputSchema.swift` in AIProxySwift to experiment with the various params. Here is a basic conversion that I used as a proof of concept, taking my twitter avatar as input with the template prompt "a cyberpunk with natural greys and whites and browns":

Snippets for you to copy and paste into your iOS and macOS apps are available in the PR that added support. Have fun!