I am admittedly kind of a noob at this, but if you enjoy my humble work, any donations are greatly appreciated: https://ko-fi.com/insult_to_ninjary
Your generosity helps me stay motivated to keep improving and to keep trying new things!
At a Glance:
Style: Anime with a small hint of realism
Works with LoRAs?: Maybe? (but not super well from my testing)
Use a VAE?: Yes! (it makes a HUGE difference)
Clip skip: 1 or 2 (2 seems to be better, but experiment with both!)
Hires. fix: use non-default upscaler; denoising strength 0.5 - 0.4
Description:
This is my first published checkpoint merge! It's just a simple merge of my AnimeReal LoRA and the Anything V4.5 base model, which was one of the best base models to use with the LoRA. I made this merge partly because I wanted to be able to use my AnimeReal model as a base model, and partly just as practice merging a LoRA into a checkpoint, which I had never done before.
In the end, I don't think this checkpoint is very useful, unless you really like the style and you're willing to put up with the issues it has. In general you're probably better off just using the AnimeReal LoRA because A) you can choose your base model for better flexibility, and B) you can adjust the LoRA weight if you're getting bad outputs. By using this checkpoint version, you're essentially stuck using the LoRA with Anything V4.5 at a LoRA weight of 1.
Other than that, this checkpoint suffers from a lot of the same problems that the LoRA version does, namely bad anatomy (especially hands and fingers), and some smudgy elements that don't have great definition. For other issues, see below.
Troubleshooting:
Most importantly, to correct poor image output that is consistent, make sure you are using a VAE. I don't 100% remember, but I'm pretty sure I did not bake a VAE into the checkpoint, so you'll have to manually use one, and it can make a HUGE difference in quality! (colors, anatomy, etc.)
From my testing, this checkpoint doesn't work great with LoRAs. You can see that the LoRAs are having an effect, but usually they further degrade the already shaky coherence of the base model. So feel free to try, but it might get frustrating!
In general, the biggest issue you're going to come across is problems with anatomy - hands, fingers, etc. Also you'll likely see some smudgy results that are not cleanly defined and coherent. Just try your best with all the usual fixes - different prompting, use a VAE, try different clip skip values, use adetailer, etc.
Outro:
If you make anything awesome, please consider posting them to the gallery! That goes for all of my models! I really like seeing what kind of stuff you all are able to make!
If you have any questions, feel free to ask! I will try my best to help! Also, if you have any ideas for future models, please don't hesitate to share them with me!