For best outputs and usage:

Sampler: DPM++ 2M Karras

Steps: 25 ~ 35, cfg 7 ~ 10, CLIP skip 2.

VAE: sdxl_vae.safetensors (mandatory)

Usage: this is a versatile type model, which means, besides to be pony, it's able to create realism (yet not photorealism), 3D and anime at once, tuned up beyond the native pony skills. To achieve the different styles:

Realistic:

Pos prompt: start it with "score_9, score_8_up, score_7_up, rating_explicit, masterpiece, best quality, amazing absurdres, perfect composition, beautiful, award winning perfect anatomy, best resolution, ultra detailed, full of details" then end it with "raw, photo, realistic, , ", being the LoRa's "2.5DRealistic" and "add-detail-xl" included in my realistic example runs.

Neg prompt: "wrinkles, furry, pony, lowres, milf, old, source_anime, anime style, low res, very lowres, rating_safe, asymmetrical face, BeyondBetaSDXLv4, EasyNegativeV2, unaestheticXL2v10", being the embeddings "BeyondBetaSDXLv4", "EasyNegativeV2" and "unaestheticXL2v10" included in my realistic example runs.

Overall note about the realistic outputs: like any pony merge, the realistic outputs may not listen as well as neither have as powerful NSFW skills as the non-realistic outputs. Make sure you're precise at depicting.

Anime:

Pos prompt: start it with "score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, source_anime" and end it with " ", being the LoRa "Smooth Anime 2 Style SDXL_LoRA_Pony Diffusion V6 XL" included in my anime example runs.

Neg prompt: "low res, very lowres, source_realistic, monochrome, 3d, render, cg, pixar, volumetric, cinematic, realism, photorealistic, easynegative, unaestheticXL_hk1", being the embeddings "EasyNegativeV2" and "unaestheticXL2v10" included in my anime example runs.

3D:

Pos prompt: start it with "score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, rating_explicit, source_3d" and end it with " ", being the "3D_Animation_Diffusion_Pony_style" LoRa included in my 3D example runs.

Neg prompt: "low res, very lowres, source_realistic, monochrome, photorealistic, source_anime, anime style, bad anatomy, old".

Note about the LoRa's and embeddings: these tools are only to enhance the model's native versatility to its max, being the model perfectly able to achieve the 3 styles without them.

Checkpoint Merge.