Caution: This is categorized as VAE, as we have no dedicated text encoder category.
Clip-L fine-tune released by zer0int on huggingface
https://huggingface.co/zer0int/CLIP-GmP-ViT-L-14
figured this might interest some due to the currently running Text-tacular Showdown Contest
TEXT: better especially for images containing text
SMOOTH: can be better for textfree images
to zer0int: in case you want this to be taken down or transferred to your own account, please contact me via direct message here onsite or on our discord
Description
sometimes better for textless images
FAQ
Comments (9)
on which folder goes this file?
it goes to your comfyui_folder/models/clip
its for sure a stupid question but I somehow did not find an "easy" answer elsewhere:
can clip-l be paired with any SDXL model?
using dual clip loader, load clip_g and clip_l and you are done.
can clip-l be paired with any SDXL model?
no
can this be used on forge?
It can be used anywhere where you can use your own CLIP-L models. Presumably even Forge, if it has some "CLIP" folder.
move downloaded files to your forge directory inside models/text_encoder and in forge you select the text encoder in the dropdown menu called vae/text encode, right next to where you select the base model.
@alphy thanks : )
Details
Files
clipLFineTuneByZer0int_smooth.safetensors
Mirrors
ViT-L-14-BEST-smooth-GmP-TE-only-HF-format.safetensors
clipLFineTuneByZer0int_smooth.safetensors
ViT-L-14-BEST-smooth-GmP-TE-only-HF-format.safetensors
ViT-L-14-BEST-smooth-GmP-TE-only-HF-format.safetensors
clipLFineTuneByZer0int_smooth.safetensors
ViT-L-14-BEST-smooth-GmP-TE-only-HF-format.safetensors
ViT-L-14-BEST-smooth-GmP-TE-only-HF-format.safetensors
ViT-L-14-BEST-smooth-GmP-TE-only-HF-format.safetensors
ViT-L-14-BEST-smooth-GmP-TE-only-HF-format.safetensors