Diffusion Single File
comfyui

Can't upscale a batch of multiple images through refiner pipeline?

#109
by synta - opened

Hey,

whenever I try to upscale a multi image batch (2+) and let it go through a second round of vae encode - ksampler - vae decode, i end up with only the first image of the batch and nothing else saved. This only happens with Anima. If I dont do it through ksampler refinement and just attach the save image node to the upscale image by node then it works. Something gets butchered through the vae encode when batch gets returned to latent space. Tried vae encode tiled, same issue.

Thanks.

Several people have confirmed the issue on their machines too.
The VAE encode node for some reason picks up only the first latent, not the one for the second or more latent images.

Hi, this isn't the ComfyUi github, please report the issue there and follow proper reporting guidelines (include workflow etc.). There is absolutely nothing that can be done for you here, yes, even if the issue is exclusive to Anima, it's up to comfy to fix it.

Hi, this isn't the ComfyUi github, please report the issue there and follow proper reporting guidelines (include workflow etc.). There is absolutely nothing that can be done for you here, yes, even if the issue is exclusive to Anima, it's up to comfy to fix it.

Okay. A workaround in the meanwhile is this node. The batch then gets into latent image by image instead of being processed as a batch. This fixed it for now.
fixbatching

Sign up or log in to comment