ModelsLab/Mangled Merge XL - V2.0

I am happy to introduce Mangled Merge XL version 2.0 to the Civitai community. This new version includes 1200 loras that have been carefully merged in to "train" the model. That's 380 more than the previous Dark version! Mangled Merge XL 2.0 also includes Hillside, Clarity, Colorful, Realistic Stock Photo, Pixelwave, Realism from Hades, Pilgrim 2D, and Crystal Clear.Usage:Mangled Merge XL does well with a CFG range between 2 and 7. There are some tokens that look much better with a low CFG score, but you can also go between 5 and 7 for the model to follow your prompt more closely if it is more complicated.The model tends to stick more to realism but can do paintings and illustrations, it's just a little harder to get them in it's current state. I want to balance it out more but I thought 1200 was a nice number to release the model with. There was/is no goal to make this model more one way or another as I am mostly working towards prompt adherence and aesthetics.Mangled Merge XL 2.0 also likes full body type long shots. It does close up portraits but if you don't specify, it will tend to output closer to full body or further away pictures. Luckily, eyes and faces come out great for this model with distance shots. Hands are decent but not perfect. But you can always touch up with some inpainting.If you are going to use strong weights, I suggest lowering the CFG or else outputs may look a little fried. Also "fur" and "1girl" can give weird outputs. Fur needs a low CFG and 1girl can potentially give unreliable outputs. It's better to use "a woman" or something along those lines. The model is prompt sensitive so it will output what is in your prompt and for the most part it doesn't output what's not in your prompt. So if you don't specify a background, it is likely to give you a white or otherwise plain background.Disclaimer:If you are trying to reproduce the example images I have provided, I am using the AI Diffusion extension for Krita which has some generation options that aren't available to add to the Civitai prompt information. Civitai only allows whole numbers for in the CFG scale inputs, but some of these models are 6.5 or 5.2 or something like that, so you may need to dial the CFG scale in to get the exact output. Civitai also doesn't have an option for the sampler I used which is Turbo/Lightning Merge - DPM++ SDE which requires the 8-steps lora. All were produced with 12 steps rather than 8 however because I thought it gave better outputs. All images were also upscaled with an additional refining process.I think that covers about everything. Have Fun!