Pure CLIP (no T5) guidance, but how ?
i tried to run workflow with single clip but could not find a way to.
i installed node LongClipTextEncodeFlux, but it also has a clip input widget which i dont know where to connect.
Can you kindly show the way to use single clip loader for flux in comfyui ?
Thank you
i have found a workflow with a single clip loader but the very left side (red marked) is not clear where its connected
anyway i used dual clip loader and loaded same clip file. so i am not using t5xxl input. if anybody wants to try.
Guidance lower than 4 gave me not very fine results, i am good with "4" if anybody wants to try their prompts.
Ah, yeah, that's not the same as just zeroing the tensor for T5; I made a node for this: https://github.com/zer0int/ComfyUI-Nuke-a-Text-Encoder
I also have two workflows, one for "CLIP only, no T5" and one for "individual CFG with CLIP, and with T5" here: https://github.com/zer0int/CLIP-fine-tune-registers-gated/tree/CLIP-vision/ComfyUI-workflows
Hope that helps! :)
Ah, yeah, that's not the same as just zeroing the tensor for T5; I made a node for this: https://github.com/zer0int/ComfyUI-Nuke-a-Text-Encoder
I also have two workflows, one for "CLIP only, no T5" and one for "individual CFG with CLIP, and with T5" here: https://github.com/zer0int/CLIP-fine-tune-registers-gated/tree/CLIP-vision/ComfyUI-workflowsHope that helps! :)
Thank you so much, i was pretty sure my workflow performed good "until" your answer revealed --> it was not the same as zeroing the tensor :))