Home Artists Posts Import Register

Downloads

Content

sdxl requires xformers, so please use venv install guide for local windows install.

Guide: https://github.com/Sxela/WarpFusion#local-installation-guide-for-windows-venv

Changelog:

v0.18.2
3.08.2023:

  • add sdxl_base support
  • temporarily disable loras/lycoris/init_scale for sdxl
  • bring back xformers :D

v0.18.3:
4.08.2023

  • add refiner support

v0.18.4:
5.08.2023

  • add weighted keywords support for sdxl
  • clear gpu vram on render interrupt

v0.18.5:

  • add sdxl lora support
  • fix load settings file = -1 not getting latest file
  • cutoff prompt at 77 tokens

v0.18.8:
10.08.2023

  • refactor lora support
  • add other lora-like models support from automatic1111
  • fix loras not being unloaded correctly

12.08.2023: moved to nightly/L tier

v0.18.9: 14.08.2023

  • disable deflicker scale for sdxl

5.09.2023:

  • moved to M tier

Basic sdxl support for sdxl_base model

Most features work, like latent scale, cc masked diffusion.

No controlnets yet, though there are some already available.

Doesn't work with embeddings yet

The results are meh, need to try refiner model, as it seems to work much better with img2img in Comfyui.

Usage:
define SD + K functions, load model -> model_version -> sdxl_base

I'd suggest trying non-vanilla XL checkpoints like dreamshaperXL, seems to work a bit better both in img2img and text2img mode (vanilla may look like a bad realesrgan upscale without further refiner pass)

You'll also have to remember the old non-controlnet settings :D

style_strength [0.5,0.3] or something.

Sdxl refiner

define SD + K functions, load model -> model_version -> sdxl_refiner

Same imitations apply, but works better as img2img, also better with lower resolutions.

Extra networks

Add other lora-like models support from automatic1111. You can now mix different lora-like networks in one run, like lora+lycoris, the call syntax is the same as before, for all types of networks - <lora:filename:weight>
Also fixed loras not being unloaded when removed from prompt.

Local install guide:
https://github.com/Sxela/WarpFusion/blob/main/README.md

Tutorials:
https://youtu.be/HkM-7wxtkGA
https://www.youtube.com/watch?v=FxRTEILPCQQ
https://www.youtube.com/watch?v=wqXy_r_9qw8
https://www.youtube.com/watch?v=VMF7L0czyIg
https://www.youtube.com/watch?v=m8xaPnaooyg

Youtube playlist with settings:
https://www.youtube.com/watch?v=wvvcWm4Snmc&list=PL2cEnissQhlCUgjnGrdvYMwUaDkGemLGq

For tech support and other questions please join our discord server:
https://discord.gg/YrpJRgVcax

Discord is the preferred method, because it is nearly impossible to provide any decent help or tech support via Patreon due its limited text formatting and inability to add screenshots or videos to comments or DMs.
Error reports in comments will be deleted and reposted in discord.

Comments

Matt Jenkins

Is it possible to run locally on a Mac versus PC/Windows (and is there an installation guide)? I'm running an AMD Radeon Pro 5700 XT with 16GB of VRAM.

Jianxun Gao

NameError Traceback (most recent call last) in () 64 65 if animation_mode == 'Video Input': ---> 66 postfix = f'{generate_file_hash(video_init_path)[:10]}_{start_frame}_{end_frame_orig}_{extract_nth_frame}' 67 if flow_video_init_path: 68 flow_postfix = f'{generate_file_hash(flow_video_init_path)[:10]}_{flow_extract_nth_frame}' NameError: name 'generate_file_hash' is not defined

Nate Denton

does anyone have a link to this to add to by models fodler. Tried using it and it of course had a melt down! sdxl_base?!?! Thanks ad sorry for being a bit of a DIV. Looked on Civit and not sure/!?

sxela

Can you describe what error you are having and at which cell?