Illustrious XL1.0 | Improved | Uncensored
I am working on fine-tuning the Illustrious 1.0 base model, and my work is not done yet! I am still making improvements, and this is the second version of the model. Let me know if you find any issues that need to be addressed!
If you want to try this model before downloading it, it is available for free on TensorArt. Here is the link: https://tensor.art/models/844678883228871126
Recommended Parameters:
Sampler: Euler A
Size : 768x1024
Steps: 25 ↔️ 30
CFG Scale: 2.0 ↔️ 6.0
Clip Skip: 2
Upscaler : R-ESRGAN 4x+ Anime6B
🌐 Any inquiries about my model are more than welcome! Feel free to reach out on Discord ➡️ shopon_skp. Let's keep the conversation going!
Description
FAQ
Comments (6)
What did you focus on improving? Backgrounds, Anatomy, Base style, the Text Encoder, etc.
Finally someone has create something like that model, the most cool thing about illustrious xl 1.0 model is the natural language understanding, unfortunately the creator is a such bad people to make the things exclusive just for a specific website...(Tensor Art). But I'm happy that we have a new people to make the things right here. Have a great development OP 🙏🖤
the devs made 1.1 open source with SDXL license , they doing great now
Interesting finetune. For some reason both original 1.0 and 1.1 don't work as intended with the macOS app I use (DrawThings).
Surprisingly your mix works. Not sure why lol.
The only noticeable downside I see is the weird fuzziness texture overlaid on the actual finished image. Not sure yet if it's the problem with either app I'm using or Loras/prompts/settings.
I'm also testing it with my own braindead mixes which tames this 'fuziness'
Overall I'm impressed with the way your model works. Keep up a good job!
it could be your vae.
this works pretty great. very sensitive on cfg.