RealCore Pony v2.0 Updates

🌍 More diverse facial features with reduced Asian bias

💡 Significantly improved lighting with elimination of unwanted overexposure

🎯 Enhanced prompt adherence and instruction following

RealCore Pony v1.0

A powerful SDXL merge built on Pony Diffusion V6 XL, blending multiple high-quality realistic models and custom LoRAs to deliver lifelike, emotionally rich, and anatomically accurate generations. Whether you're creating intimate portraits, photos, or detailed NSFW content — RealCore Pony strikes the perfect balance between realism, expressiveness, and stylistic flexibility.

📌 Note: All example images and videos were generated using the DMD2 LoRA with the LCM sampler (4–8 steps) in ComfyUI, demonstrating the model’s speed and quality under accelerated inference.

⚠️ Important considerations:

While RealPonyCore excels in NSFW realism, it is less diverse in artistic styles and facial variety compared to RealCore XL. It prioritizes anatomical fidelity and photo-like rendering over broad stylistic range.

Like most Pony-based realism models, it struggles with low-resolution generations. To address this, RealCore Pony intentionally incorporates high-resolution-trained components and omits low-res-optimized layers.

Recommendation: Always use HiRes Fix (or an equivalent upscaling + detail refinement pass) for best results.

You can also take a look at the Huggingface model repository, there is an fp32 version there
👉 RealCore Pony on CivitAI

🧩 ComfyUI Workflows

📥 Download LCM DMD2 Workflow

⚙️ To use KSamplerWithNAG, install the ComfyUI-NAG custom node:

cd ComfyUI/custom_nodes  
git clone https://github.com/xmarre/ComfyUI-NAG.git

Для русскоязычных пользователей

Если вы интересуетесь нейросетями, LLM, Stable Diffusion и хотите пообщаться в тёплой компании, присоединяйтесь к нашему ламповому чату:
👉 @ritya_smokeroom

Description

🧾 Checkpoint Recipe

version 0.1.0
model "ritya\\RealCore_Pony_v10.safetensors" model_config="sdxl-sgm" merge_space="weight"
model "pony\\HanfuPonyNSFW_ponyV3.safetensors" model_config="sdxl-sgm" merge_space="weight"
model "pony\\bigLove_pony3.safetensors" model_config="sdxl-sgm" merge_space="weight"
model "pony\\damnIllustriousPony_v30OUTDATED.safetensors" model_config="sdxl-sgm" merge_space="weight"
merge "🔨_karcher_mean_with_json" &1 &2 &3 json_params="{
  \"global\": \"0.234,0.383,0.383\",
  \"max_iter\": 30,
  \"tol\": 3e-7,
  \"OUT04\": \"0.333,0.226,0.441\",
  \"OUT05\": \"0.204,0.301,0.495\",
  \"OUT06\": \"0.333,0.226,0.441\",
  \"OUT08\": \"0.231,0.385,0.385\"
}"
dict BASE=1.0 IN00=0.0 IN01=0.0 IN02=0.0 IN03=0.0 IN04=0.0 IN05=0.0 IN06=0.0 IN07=0.0 IN08=0.0 M00=0.0 OUT00=0.0 OUT01=0.0 OUT02=0.0 OUT03=0.25 OUT04=0.5 OUT05=0.75 OUT06=0.5 OUT07=0.5 OUT08=0.5 VAE=0.0
literal &5 model_config="sdxl-supermerger_blocks" merge_space="param"
merge "convert_sdxl_blocks_to_sgm" &6
merge "weighted_sum" &0 &4 &7
model "pony\\realistic\\RSGRL-CR-PONY(10_10_25).safetensors" model_config="sdxl-kohya_kohya_lora" merge_space="weight"
merge "convert_'sdxl-kohya_kohya_lora'_to_base" &9
merge "convert_'sdxl-kohya'_to_'sdxl-sgm'" &10
model "pony\\realistic\\lora.TA_trained.safetensors" model_config="sdxl-kohya_kohya_lora" merge_space="weight"
merge "convert_'sdxl-kohya_kohya_lora'_to_base" &12
merge "convert_'sdxl-kohya'_to_'sdxl-sgm'" &13
model "pony\\realistic\\BSS_NEOSTR.safetensors" model_config="sdxl-kohya_kohya_lora" merge_space="weight"
merge "convert_'sdxl-kohya_kohya_lora'_to_base" &15
merge "convert_'sdxl-kohya'_to_'sdxl-sgm'" &16
model "pony\\realistic\\BSS_CNMTK-R(Restored).safetensors" model_config="sdxl-kohya_kohya_lora" merge_space="weight"
merge "convert_'sdxl-kohya_kohya_lora'_to_base" &18
merge "convert_'sdxl-kohya'_to_'sdxl-sgm'" &19
merge "🔨_ties_lora_with_json" &11 &14 &17 &20 json_params="{
  \"lambda\": 1.0,
  \"weights\": \"0.400,0.350,0.200,0.150\",
  \"density\": 0.4,
  \"vote_sgn\": true,
  \"apply_stock\": false,
  \"apply_median\": false
}"
literal 1.0 model_config="sdxl-sgm" merge_space="param"
merge "pick_component" &22 "clip_l"
merge "pick_component" &22 "clip_g"
merge "fallback" &23 &24
literal 0.0 model_config="sdxl-sgm" merge_space="param"
merge "pick_component" &26 "vae"
merge "fallback" &25 &27
merge "pick_component" &22 "diffuser"
merge "fallback" &28 &29
merge "add_difference" &8 &21 &30
merge "fallback" &31 &8
merge "add_cosine_a" &0 &32 0.5
merge "pick_component" &26 "diffuser"
merge "fallback" &28 &34
merge "weighted_sum" &33 &4 &35