Sign In

ControlNetXL (CNXL)

1.8k
88k
275
Updated: Mar 4, 2024
toolcontrolnetsdxlcnxl
Type
Checkpoint Trained
Stats
1,215
Reviews
Uploaded
Dec 19, 2023
Base Model
SDXL 1.0
Hash
AutoV2
9BE9428991
0
0
0
0

bdsqlsz: canny | depth | lineart-anime | mlsd-v2 | normal | openpose | recolor | segment | segment-v2 | sketch | softedge | t2i-color-shuffle | tile-anime-α | tile-anime-β | NEW: tile-real
briaai-canny || CVL-Heidelberg: canny | depth || destitech: inpaint
diffusers: Canny small | mid | full, Depth small | mid | full | zoe
h94: ip-adapter | ip-adapter-vit-h | ip-adapter-plus | ip-adapter-plus-face
kohya-ss: real blur | canny | depth
└➤ anime blur | blur-beta | canny | depth | openpose | openposev2 | replicate | replicatev2 | scribble
Stability.ai: canny | depth | recolor | revision | sketch
SargeZT: depth-16b-zoe | depth-faid-vidit | depth-zeed | depth-zoe | softedge
└➤ t2i adapters binary | canny | color | depth | segmentation | softedge
TencentARC: canny | depth-midas | depth-zoe | lineart | openpose | recolor | sketch
NEW: ttplanet: tile-real || thibaud: openpose | openpose-lora

ControlNetXL (CNXL) - A collection of Controlnet models for SDXL

Good news everybody - Controlnet support for SDXL in Automatic1111 is finally here!

This collection strives to create a convenient download location of all currently available Controlnet models for SDXL. If you are the author of one of these models and don't want it to appear here, please contact me to sort this out. The naming scheme of the files follows lllyasviel's from here https://huggingface.co/lllyasviel/sd_control_collection/tree/main.

Requirements for Automatic1111: at least version 1.5.0, better upgrade to the latest version of 1.6.0. + sd-webui-controlnet extension v0.400+. This is all still very new, so bugs or weird behaviour might occur. If you encounter any irregularities you can join us on our discord and ask for support or get in contact with the developers of Automatic1111/Controlnet via github/huggingface.

Known issues:

  • kohya-ss' controlnet lllite models seem to only have an effect when used with resolutions that are dividable by 16. supposed to be fixed in 1.1.410

  • currently there is no preprocessor for the blur model by kohya-ss , you need to prepare images with an external tool for it to work. there is now a preprocessor called gaussian blur.

  • As files with the extension .pth can't be uploaded the ip-adapter.pt from h94 has to be renamed manually after downloading. Reuploaded as .bin/.safetensors.

  • The two versions of the control-loras from Stability.ai are marked as fp32/fp16 only to make it possible to upload them both under one version.

  • There have been reports especially for the control-loras to sometimes produce garbage when using scaled-dot-product cross optimization, xformers seems to be the better choice for them.

  • Deforum + Controlnet seems to break with the latest update of Automatic1111. Until a fix arrives you can downgrade to 1.5.2. seems to be fixed with latest versions of Deforum and Controlnet extensions.

A huge thanks to all the authors, devs and contributors including but not limited to: the diffusers institution, h94, huchenlei, lllyasviel, kohya-ss, Mikubill, SargeZT, Stability.ai, TencentARC and thibaud.