And here it is! 🥳🎊🍺
As promised, here the naive converter I built to leverage the massive SD1.5 list of Textual Inversion with SDXL models: https://github.com/nArn0/sdxl-embedding-converter
And here is for example the result when converting this TI:
woman, wendyGNRC, sfw, black lace midi dress, holding a metallic clutch, cowboy shot, in a hotel lounge
(Generated using UnNamedIXL)
The Textual Inversion is correctly loaded as seen using Forge:
[Textual Inversion] Used Embedding [wendyGNRC] in CLIP of [clip_l]
[Textual Inversion] Used Embedding [wendyGNRC] in CLIP of [clip_g]Using this converter assume you have a bit of knowledge regarding installing python, pytorch and using a command line tool, but really, it is not as scary as it seems 😁
If you already have A1111/Forge/Reforge/ComfyUI/etc... installed you probably already have python virtualenv containing the needed packages (torch and tqdm) so, you should be able to use the converter in the corresponding folder.
PS: for anyone not familiar with Github, the safetensors model file is in the release.