Name: Ultra Realistic By Stable Yogi (ILLUS)
Rating: 5 (1369 reviews)
Author: Stable_Yogi

Full ComfyUI in your browser, no GPU needed. Pay Buzz per second only when you run. Now in preview. [Open Comfy on Civitai](https://comfy.civitai.com)

Comfy on Civitai

Loving this model? <a target="_blank" rel="ugc" href="https://forgebun.com">Get the Pro version of my models here</a> for Pro versions of my checkpoints, Lora's and other resources.To discuss custom Lora's or models, feel free to connect on <a target="_blank" rel="ugc" href="https://discord.gg/FAgfAeMqys">Discord.</a><h3 id="ultra-realistic-by-stable-yogi-(illus)">Ultra Realistic By Stable Yogi (ILLUS)</h3>Ultra Realistic by Stable Yogi merges the photoreal skin, fabric and lighting of Realism_By_Stable_Yogi with the flattering poses, proportions and vibe of Babes By Stable Yogi into one easy-to-use checkpoint—so you get ultra-realistic, studio-grade results in a single download without having to switch models or stack LoRAs.For details on each version read the "About this version" section.

This checkpoint was built with a focus on beautiful, lifelike results — expressive faces, thoughtful composition, and soft, natural lighting that gives each image a polished, cohesive feel. It leans toward rich color, fine detail, and a consistent aesthetic across a wide range of scenes and moods. A lot of care went into getting the look just right, and I hope it brings something special to your creations. Thanks so much for trying it out — I'd love to see what you make!The Model on this page. fp16 - Normal version Q4 GGUF Fp8

v2.5

This checkpoint was built with a focus on beautiful, lifelike results — expressive faces, thoughtful composition, and soft, natural lighting that gives each image a polished, cohesive feel. It leans toward rich color, fine detail, and a consistent aesthetic across a wide range of scenes and moods. A lot of care went into getting the look just right, and I hope it brings something special to your creations. Thanks so much for trying it out — I'd love to see what you make!The Model is fp16 (DMD2 4 step version) Use 8 Steps CFG 1.5 LCM

v2.5 DMD2

ULTRA REALISTIC BY STABLE YOGI ILLUS V2 (FP16)What’s new in V2: Sharper faces, improved body structure, cleaner hands/fingers, richer skin texture, and more variety in facial types—while keeping realism and strong prompt adherence intact.Quickstart (Defaults)<ul><li>Sampler: DPM++ SDE Karras (default)</li><li>Steps: 27 (min 20)</li><li>CFG: 4–5 (works great 3–7)</li><li>Resolution: All SDXL</li><li>ADetailer: Recommended</li><li>High-Res Fix: Optional</li><li>Upscale: ≥ 1.5× with 4x-UltraSharp</li><li>VAE: Auto (included)</li></ul>What is FP16?<ul><li>Standard build for maximum compatibility across GPUs.</li><li>Balanced detail, stability, and speed.</li><li>Recommended for most users.</li></ul>

v2.0_FP16

ULTRA REALISTIC BY STABLE YOGI ILLUS V2 (BF16)What’s new in V2: Sharper facial realism, improved anatomy, stronger hands/fingers, richer skin texture, and enhanced lighting depth—optimized for cutting-edge hardware.Quickstart (Defaults)<ul><li>Sampler: DPM++ SDE Karras (default)</li><li>Steps: 27 (min 20)</li><li>CFG: 4–5 (works great 3–7)</li><li>Resolution: All SDXL (works perfectly at 896×1152 and higher)</li><li>ADetailer: Recommended</li><li>High-Res Fix: Optional (for large outputs)</li><li>Denoise: 0.30</li><li>Hi-res steps: ≥ 5</li><li>Upscale: ≥ 1.5× with 4x-UltraSharp</li><li>VAE: Auto (included)</li><li>Schedule type: Automatic or Karras</li><li>Compatible samplers: DPM++ 2M SDE, Euler a, DPM2 a</li></ul>What is BF16?<ul><li>Optimized for newest GPUs &amp; datacenter cards (e.g., RTX 6000 Ada, A100, H100, RTX 5090).</li><li>Delivers higher precision and efficiency when BF16 is supported.</li><li>Use only if your hardware supports BF16 natively; otherwise pick FP16.</li></ul>

v2.0_BF16

ULTRA REALISTIC BY STABLE YOGI ILLUS XL V2 (DMD2)What’s new in V2: Lightning-fast generation with sharper faces, improved hands/fingers, and more realistic body structure—while staying ultra-efficient at low steps.Quickstart (Defaults)<ul><li>Sampler: LCM, Euler A (default)</li><li>Steps: 6 (base)</li><li>CFG: 1 (works best between 1–1.5)</li><li>Resolution: All SDXL</li><li>ADetailer: Recommended</li><li>High-Res Fix: Optional (for large outputs)</li><li>Denoise: 0.30</li><li>Hi-res steps: ≥ 5</li><li>Upscale: ≥ 1.5× with 4x-UltraSharp</li><li>VAE: Auto (included)</li><li>Compatible samplers: Euler a, LCM</li></ul>What is DMD2?<ul><li>Based on Tianweiy’s DMD2 Lightning LoRA project: <a target="_blank" rel="ugc" href="https://github.com/tianweiy/DMD2">https://github.com/tianweiy/DMD2</a></li></ul>

v2.0_DMD2

ULTRA REALISTIC BY STABLE YOGI ILLUS V2 (DMD2 ALT)What’s new in V2: All the speed of DMD2, but tuned for higher CFG, more steps, and richer detail. Faces, bodies, and lighting hold up even when pushed to heavier prompts.Quickstart (Defaults)<ul><li>Sampler: DPM++ SDE Karras (default)</li><li>Steps (base): 6 at CFG 1</li><li>CFG: 1–7 supported</li><li>Scaling rule: For each +1 CFG, add ~2 steps (e.g., CFG 2 → 8 steps, CFG 3 → 10 steps, … up to CFG 7 → ~30 steps)</li><li>Resolution: All SDXL</li><li>ADetailer: Recommended</li><li>High-Res Fix: Optional (for large outputs)</li><li>Denoise: 0.30</li><li>Hi-res steps: ≥ 5</li><li>Upscale: ≥ 1.5× with 4x-UltraSharp</li><li>VAE: Auto (included)</li><li>Compatible samplers: DPM++ SDE Karras, DPM++ 2M SDE, Euler a, DPM2 a</li></ul>What is DMD2 Alt?<ul><li>A finetuned expansion of DMD2 by Stable Yogi.</li><li>Retains the lightning-fast generation core, but scales up gracefully with higher CFG and longer steps.</li><li>Ideal if you want DMD2 speed with FP16-level detail and prompt adherence.</li><li>Stronger anatomy, cleaner faces, and better compositional control at medium–high CFGs.</li></ul>

v2.0_DMD2_ALT

<ul><li>Model: Ultra_Realistic_By_Stable_Yogi_ILLUS_V1_CIV_FP16 (initial version)</li><li>Steps: 27</li><li>Samplers: Euler a, DPM++ 2M SDE</li><li>Schedule type: Automatic or Karras</li><li>CFG scale: 5</li><li>Resolution: 896 × 1152</li><li>Style mix: combines “Babes” poses/proportions with hyper‐detailed “Realism” textures</li><li>Strengths: outstanding detail in anatomy and hands; rich colours, contrast, and lighting</li><li>Usage: ideal for all prompt types and image subjects</li><li>Feedback: I’m open to your observations—please share any comments!</li></ul>

v1.0_FP16

<ul><li>Model: Ultra_Realistic_By_Stable_Yogi_ILLUS_V1_CIV_DMD2 (initial version)</li><li>This is a Lightning Model based on DMD2 generates images in just 6 steps.</li><li>Steps: 6</li><li>Samplers: Euler a, LCM</li><li>Schedule type: Automatic or Karras</li><li>CFG scale: 1</li><li>Resolution: 896 × 1152</li><li>Style mix: combines “Babes” poses/proportions with hyper‐detailed “Realism” textures</li><li>Strengths: outstanding detail in anatomy and hands; rich colours, contrast, and lighting</li><li>Usage: ideal for all prompt types and image subjects</li><li>Feedback: I’m open to your observations—please share any comments!</li></ul>

v1.0_DMD2

<ul><li>Model: Ultra_Realistic_By_Stable_Yogi_ILLUS_V1_CIV_BF16 (initial version)</li><li>Steps: 27</li><li>Samplers: Euler a, DPM++ 2M SDE</li><li>Schedule type: Automatic or Karras</li><li>CFG scale: 5</li><li>Resolution: 896 × 1152</li><li>Style mix: combines “Babes” poses/proportions with hyper‐detailed “Realism” textures</li><li>Strengths: outstanding detail in anatomy and hands; rich colours, contrast, and lighting</li><li>Usage: ideal for all prompt types and image subjects</li><li>Feedback: I’m open to your observations—please share any comments!</li></ul>