EquinePotpourri

V2 Message.

Well, he's gone and done it folks. I've been like a dog with a bone for a while now. I'd like to present to you my newest merge. The 1/32 merge, that includes 20 models and their clips merged by two at 50%, forking into the next layer of the merge.  There were 5 layers to the merge, combining everything by 2 at 50%. I added detailers to the models that I felt needed a little help. In the end I ended up disabling the lora's because it was concentrating to hard on faces, making it almost impossible to get anything else out of this new model. All models were collected from here on civitai. Some of them have been removed for some reason or another. No idea why. I’m fairly happy with how the model if feeling at the moment, but I've done the leg work for v3 in the workflow, but adding the lora in already. The workflow is attached with the checkpoint, so just drop if into comfy if you're interested in what when into it.

There were some models added that weren't pony, but the majority of them were. For this reason, using things like "loli, child, young," in your negative might be a necessity. Well that's it for now cause I'm sure many of you will think, TL'DR.

I hope you enjoy using EquinePotpourri-V2. 😜It being a mix of just about everything I felt the name was fitting.

I'd tried finding a lora to merge in with this that'll allow for selfsucking gens until I realized that it's a pony model and all I needed to do was to use the right words and It'd probably give me what I was asking for.  "Rating_ecplicit (self oral), (self suck)," was what worked for me. I figured I'd share my discovery If anyone is interested.

VAE is included in the model.

---

V1 Message.

It's a pony model, So it'll do everything a pony can do. I've merged with models that are lifelike or realistic, but adding 'source_anime,', or one of the others to your prompt gives some interesting results. I also had some pretty quick results with some lcm loras too, 'LCMV2-PONYplus-Paseer' https://civitai.com/models/216590/paseer-sdxlpony-lcm-and-turbo-accelerator is my go to at the moment. Using the LCM Sampler and Turbo Scheduler with that lora loaded while in SwarmUI is super quick.

I haven't filtered anything out of this merge. So I recommend checking out the Negative prompt section.

Well that's it from me for now, Happy prompting. Would love to see what you guys and gals are able to come up with.

Positive prompt:

score_9, score_8_up, score_7_up, score_6_up, source_anime, source_pony, source_furry, source_cartoon, rating_questionable, rating_explicit, Photoreal, masterpiece, extremely detailed, finely detail, ultra-detailed,

Negative prompt

score_5, score_4,  child, less than 18yo,

Usage

Used CFG Scale 7, Sampling steps 35, and 1024x1024 Clip skip 1

 but my 3060 prefers less.  CFG between 4 - 7 is all I've tried.

4x_foolhardy_Remacri was used to upscale the images.