What API design would you like to have changed or added to the library? Why?
img2img currently uses the init_image parameter, whereas inpaint uses the image parameter.
What use case would this enable or better enable? Can you give us a code example?
I have some code that dynamically switches between StableDiffusionPipeline, StableDiffusionImg2ImgPipeline, and StableDiffusionInpaintPipeline based on whether an image and mask are provided, and standardizing on init_image would simplify the pipeline call (instead of checking beforehand / supplying unused parameters):
pipeline(prompt, init_image=image, ...)