libwaifu-wai-illustrious-v17

WAI-illustrious-SDXL v17.0 for libwaifu

WAI-illustrious-SDXL v17.0 converted to the format libwaifu reads, so it can draw from it directly. The weights are the original ones: the U-Net and both text encoders are stored as float16, and the VAE is left wider because it overflows float16.

Converted from the author's own release, WAI-illustrious-SDXL v17.0 by WAI0731, file waiIllustriousSDXL_v170.safetensors, sha256 f116b0c78ff441467b0cdc8f1936e1ed18ea31e9997c7b132b1b8db533f0bd04 -- the hash the author publishes, so these weights are verifiably theirs. It is an Illustrious XL fine tune, so it is prompted with danbooru tags rather than with sentences.

This is an epsilon-prediction model, which is what libwaifu's Euler sampler reads. Both halves of the autoencoder are here, so it draws from a picture as well as from a prompt.

Files

A model is a .yaml and the files it names, all of them in one directory. The manifest is a small text file you can read and edit; the weights are ordinary safetensors, readable by torch, numpy or the viewer on this page; the tokenizer is the tokenizer.json Stability published.

The split is between tensors, never inside one, so each file is a safetensors in its own right and can be read and checked alone. The picture the four draw is identical, to the byte, to the one an unsplit model draws.

Usage

waifu knows this model as sdxl:wai, so there is nothing to download by hand:

waifu draw -m sdxl:wai

sdxl:wai:v17 names this release and keeps meaning it; sdxl:wai follows whatever is current.

hf download ling0322/libwaifu-wai-illustrious-v17 --local-dir wai-illustrious-v17
waifu draw -m wai-illustrious-v17/wai-illustrious-v17.yaml

Keep the files side by side -- a manifest names its neighbours by file name and will not follow a path anywhere else.

Prompting

This model wants fewer tags, not more. Its author asks that quality and aesthetic tags not be piled on and that negative prompts be kept short, because doing otherwise makes the picture worse -- which is the opposite of what most Illustrious fine tunes ask for. The prompt the manifest suggests carries no quality prefix for that reason.

Conversion

python tools/sdxl_exporter.py \
  -checkpoint waiIllustriousSDXL_v170.safetensors \
  -output wai-illustrious-v17.safetensors \
  -part-size 2GB

License

The weights are the author's; this repository only changes the container they are in, and the terms are theirs rather than libwaifu's MIT. On the model page the author allows derivatives and does not require credit, but permits no general commercial use -- only selling the images it generates and running it on Civitai. See the model page for the terms as they stand.

About v17, from the author

• Fixed issues where some characters had inconsistent color tones with the background.
• Attempted to fix the issue where the background had low relevance to the characters.
• The overall coloring in this version is slightly smoother and cleaner than before.
• Attempted to improve the high-resolution fix's ability to correct limbs; using it has a high chance of automatically fixing issues with arms, legs, and hands/feet.
• If you're not satisfied with the default character age, try adding (aged up:1.0-2.0) or (mature female:1.0-2.0) at the beginning and adjust as needed.

Please do not add too many quality and aesthetic-related tags, nor overly long negative prompts, as this will actually reduce image quality and make it more blurry.

The weights are the author's; this repository only changes the container they are in. See the source repository for the full model card.