SD3 how exactly?

[mweldon]

Okay I assume SD3 must be working for someone, but I don't understand how. There are three models. The only one that doesn't give errors on load is "stableDiffusion3SD3_textEncoderT5E4m3fn.safetensors". So I use that. I don't know which clip models I need so I put all 4 models in a folder called models/clip. Set the VAE to None or Automatic or whatever, makes no difference. Euler, 30 steps, CFG 8, 1K x 1K. All I get is what looks like random noise.

I have not found any instructions anywhere. I'm sure I'm doing something wrong because it looks like others are getting some results, but nothing I have tried gets me anything but random noise. I know the models are fine as I have tested them in ComfyUI.

This is with the dev branch. I tested it a few weeks ago in the sd3 branch and was getting the same results. At this point I assume I'm just missing something on my end but I have no idea what.