Workflow for XL/Illustrious/NoobAI Models
Any feedback would be appreciated.
See the Article for a Guide:
https://civarchive.com/articles/17339 (probably doesn't work rn)
https://civarchive.com/articles/17339
Key Features:
LoRA
TIPO Prompting
Detail Enhancers
ControlNet Upscale
IPAdapter
HiresFix
OpenPose
Wildcards
ControlNET
Color Match
FBCNN
Regional Prompting
Inpaint
Outpaint
Workflows included:
🟥Advanced: comprehensive text to image workflow
🟨Basic: more basic version of the Standard workflow
🟦Standard: more basic version of the Advanced workflow
🟩Detailer: detailing previously generated or existing images
Experiment and enjoy!
Description
Update from V29→V30
(This is not an exhaustive changelog)
General:
General improvements and bug fixes
Adjusted default values (e.g. cfg)
Advanced:
Changed default QwenVL model in alternative workflow
Split HiresFix into Hires Pre/Post Detailer pipeline
Basic:
Split HiresFix into Hires Pre/Post Detailer pipeline
FAQ
Comments (9)
Excuse me, can it be used for multiple characters? Or do you have another workflow for that?
I have a few questions, if you can help me. And a update request at the end:
1. I feel like my MaskDetailer is generating low res results after the inpaint. Is there some settings so I can try to increase the resolution to fit my currently upscaled one? I tried messing with the Guide Size node but it changed nothing. And I don't know where I should play with to try and increase the res.
2. If I want to upscale my image like the hi-res fix from auto1111, increasing details and stuff, which one should I use? The HiRes toggle or Upscaler one?
As for the request: Can you add the Lora Loader and Trigger word toggle nodes to the detailer workflow? :3 that would be awesome!
I’m not entirely sure about the first issue, as I couldn’t observe what you described on my end.
To increase details when upscaling, use the "Upscaler" option. Alternatively, you could also use hires + detailer.
p.s. I added the Lora Loader and Trigger Word Toggle to the detailer workflow.
Everything works except for ultralytics, which can't find files or directory despite being able to detect said files in the node. I've essentially just skipped enabling it for now.
My goodness this workflow is solid. Even with 8 gb of VRAM I am generating 8k images with impeccable levels of detail. Hoping for some guidance on 2 characters with Regional Prompting in the near future. I made a mask and loaded it in. I even generated a few images with it but the results were somewhat lackluster. Maybe you can update the guide to include some pointers on that. I really appreciate the work you've been doing with this workflow. It's literally the best I've ever used and it keeps getting better as updates roll out. Thank you so much for your work.
I'm curious what the hiresfix is? Because it doesn't seem to upscale in latent space like the A1111 does, it just seems to upscale the image after it's been decoded.
HiresFix uses an upscale model (e.g. 4x_foolhardy_Remacri), to upscale the image after generation