AnonymousCervine / depth-image-io-for-SDWebui

An extension to allow managing custom depth inputs to Stable Diffusion depth2img models for the stable-diffusion-webui repo.
72 stars 6 forks source link

txt2img with custom depth input doesn't work when using "Hires. fix" feature #7

Open AnonymousCervine opened 1 year ago

AnonymousCervine commented 1 year ago

This is due to a regression in the SD webui project, not due to an error in this extension. (The same behaviour can be triggered without this extension, using a dedicated inpainting model in txt2img mode and "Hires. fix" enabled.)

Priority: Low for now. Feel free to raise your voice here if this issue affects your workflow and/or you think it needs more attention.

Workaround: Go through the approximate steps of hires-fix manually: Render, upscale, then fill in details with img2img.