This might seem trivial but this appears to be quite standard with other generators that allow init images as input, including other versions of SD and previous diffusion/older models. So I was wondering is there a reason why this was removed/disabled?
Of course it depends on an individual's workflow, but when working with init images, it is often much more convenient to work from original high resolution files.
For example I may want to take an image with 1:1 ratio and try generating with it at 512x512, 576x576, 640x640, etc. Or doing multiple small edits to init image(s) between runs. Currently it can get quite tedious having to resize and maintain multiple versions of the same images. Exponentially bothersome when working with variations or animated sequences of init images ;)
It would make things so much more simple if img2img took care of resizing to specified -W and-H values, instead of using an external app.
This might seem trivial but this appears to be quite standard with other generators that allow init images as input, including other versions of SD and previous diffusion/older models. So I was wondering is there a reason why this was removed/disabled?
Of course it depends on an individual's workflow, but when working with init images, it is often much more convenient to work from original high resolution files.
For example I may want to take an image with 1:1 ratio and try generating with it at 512x512, 576x576, 640x640, etc. Or doing multiple small edits to init image(s) between runs. Currently it can get quite tedious having to resize and maintain multiple versions of the same images. Exponentially bothersome when working with variations or animated sequences of init images ;)
It would make things so much more simple if img2img took care of resizing to specified -W and-H values, instead of using an external app.