50 lines
1.6 KiB
Markdown
50 lines
1.6 KiB
Markdown
# Negative-attention-for-ComfyUI-
|
|
Takes the difference in between the positive and negative conditioning at the cross attention.
|
|
|
|
This is an experiment.
|
|
|
|
Only tested with SDXL and SD 1.X.
|
|
|
|
Will not work with Flux (see bottom note).
|
|
|
|
This allows to:
|
|
|
|
- Get a negative influence without generating a negative prediction
|
|
- Let the unconditional prediction be unconditional
|
|
- Or doubling down by having the same done for the negative prediction (overblown results are to be expected unless using an anti-burn or low CFG scale)
|
|
|
|
In order to do this for now the negative conditioning is sneaked to the attention by being concatenated to the positive by using a special node.
|
|
|
|
|
|
|
|
They are then split at the half before the cross attention.
|
|
|
|
Like any model patcher, it is to be plugged right after the model loader:
|
|
|
|
|
|
|
|
An example workflow is provided.
|
|
|
|
## Example:
|
|
|
|
No modification:
|
|
|
|
|
|
|
|
Strength at 2, rescale after ON:
|
|
|
|
|
|
|
|
Strength at 2, rescale after OFF:
|
|
|
|
|
|
|
|
|
|
## Note
|
|
|
|
I haven't managed to make this work with anything but SDXL / SD1.5
|
|
|
|
I did spend two hours looking for how to patch the equivalent of the cross attention for Flux but did not find how (like the keywords for the patch or something).
|
|
|
|
Any help appreciated!
|