>>105572477
looks like you need to work on your technique. pony for example is notoriously bad at backgrounds, so that model would really not be your best choice for a that sort of thing. if you want help, post something that helps us understand what you are aiming for. img2img in comfy works flawlessly and so do the various controlnets. so yeah, drop an image of your 'vision' and a screencap of your workflow.
>>105572839
what a peculiar request. we talking anime?