Beast 5C SEPTEMBER FREE 九月全免费
BUZZ PayAccess Version/ NOW FREE FOR ALL
Beast 5C KREA2 TURBO 多彩青春 Colorful Youth
基于红潮赤佬5 BUZZ Deal (K2T) 商务版 T2I制作
量化 BF16/FP8S/INT8C/GGUF Q4km/4-bits 全放出
将角色年龄（倾向）调校为更适合年轻用户的阶段
结合卡通化的多彩情绪表达与更具活力的动作语言
Built with the [RedCraft 5 BUZZ Deal (K2T) Commercial Edition T2I].
The model is tuned toward a younger, more community-friendly character age range, combined with colorful cartoon-inspired emotional expression and a more dynamic, energetic visual language.
例图附带工作流，如果平台没有显示出来可以下载原图
下载列表附带例图所用的简易工作流
Example images include the workflows used to create them.
If the workflow is not displayed by the platform, simply download the original image to retrieve it.
A simplified workflow is also included with the download package.
Key Features
单模型 / 单次采样 / 8-STEPS / Euler / Krea2支持Tags提示词 / 无需PE增强
Single Model
Single Sampling Pass
8 Steps
Euler
KREA2 Tag Prompt Support
No prompt enhancement needed
Hardware
推荐6GB以上显卡使用，内存16-32G即可
Recommended: GPU with 6 GB+ VRAM
System RAM: 16–32 GB is sufficient
没有复杂的配置，没有复杂的玩法，没有学习曲线
没有耗时间的重绘，没有任何局部重修，上手就用
No complicated configuration.
No complicated tricks.
No learning curve.
No time-consuming redraws.
No local retouching or regional repainting.
Load it and start creating.
Dark Beast 黑兽H3 Director2 GS
做好工具人 服务艺术家
Dark Beast | 黑兽 H3 Director2 God speed 神速短剧生产线 09/26/2026
运营成本极低的 ?.???/s 支持 serverless 自动启停( 无单0费用 )
非开源推理框架(RED Attention 稠密/稀疏 双重算子动态切换)
支持30s以上单片生产(OVDN)，支持上下文滑动窗口连续出片
全兼容 sm_75/sm_86/sm_100/sm_120/sm_103 (自编译cu_bin)
黑兽Director2 GS因技术问题（审核中）
The "RedCraft" "Black Beast" and "REDGraft LTX" model 站内信咨询
该模型下载 Access 包含以下所有模型（及更新服务）
Recommended to purchase BUZZ
RedCraft | 红潮 | REDMIX MCP + LTX25 2k 16-bit HDR- 08/25/2026
https://civarchive.com/models/958009/redcraft-or-or-hybrid-h3-a2a-beta2-ltx25-2k
REDGraft | LTX 2.5 | 老同学 Fast 2K 16bit-HD| sulphur2 ported 移植版
https://civarchive.com/models/1295569/redgraft-ltx-25-fast-2k-or-sulphur2-ported
付费用户提供一对一配置服务
神速版本服务器编译及部署模型包
Recommended to purchase BUZZ
One-on-one configuration services are provided to paid users.
商用 PRO6000 96G 显存设备，输出2k视频≈90-100秒
商用 PRO4500/ 5090 32G 显存设备，输出 2k 视频≈100-120秒
家用 3090/4090 24G 显存设备，输出 2k 视频≈140-180秒
GB200/GB300 单核心/多核心，输出 2k 视频≈90秒 / 支持实时直播
CPU/RAM 内存需求 72-128G （15-30秒）
The "RedCraft" "Black Beast" and "REDGraft LTX" model series
Dark Beast | 黑兽 H3 Director Edition 自动短剧生产线 08/28/2026
是原生H3单次采样直出2k分辨率的高清方案（ 768P/VSR/RIFE TensorRT ）
所有示例短片均为 6-10 步 黑兽 H3 modified 版本单次采样生成
该模型下载 Access 包含以下所有模型（及后续更新服务）
RedCraft | 红潮 | REDMIX Hybrid A2A beta2 + LTX25 2k 16-bit HDR- 08/25/2026
https://civarchive.com/models/958009/redcraft-or-or-hybrid-h3-a2a-beta2-ltx25-2k
beta2 版本融合了 LTX2.5 distilled 作为2k高清化方案，并且消除了音频bug。
The beta2 version integrates LTX2.5 distilled as a 2K HD solution and has eliminated audio bugs.
REDGraft | LTX 2.5 | 老同学 Fast 2K 16bit-HD| sulphur2 ported 移植版
https://civarchive.com/models/1295569/redgraft-ltx-25-fast-2k-or-sulphur2-ported
LTX 16 bit HDR & Nvidia RTX VSR
The LogC3 transform and inverse transform are used prior and after generation (accordingly) in order to achieve the dynamic range of 16 bit within normal number range.
RIFE TensorRT Frame interpolation
工作流预览： https://civarchive.com/models/579280
以上付费用户提供一对一配置服务
One-on-one configuration services are provided to paid users.
Recommended to purchase BUZZ
Online Buzz Payment & Model Access
Online Buzz payment is supported, granting direct access to the models.
Payments in fiat currency or cryptocurrency are also accepted.
Regardless of the payment method chosen, you will gain access to all of my models.
If the service configuration requires the use of third-party models, I will pay the corresponding fees to the contributors.
Contact me via CivitAI private message
let me know your requirements and configuration needs.
Thank you！ 🙏
[NEW] Krea2-Dark-Beast3 黑兽3.0 INT8 Convrot/BF16 07/28/2026( Enhance Anatomy & Niji Distilled )
[NEW] LTX2.3 GTA 高速版 Grand Truth Animation 7/19/2026
[NEW] KREA2 GPT 逼真版 Grand PUSSY Truth 7/13/2026
Krea2RedMix3.1-INT8/INT4-Convrot-ComfyUI (visual effects)
REDZimageTurbo2.0-INT8-Convrot-ComfyUI (ZIT-HD 2026)
Uncensored, Unfiltered, Unapologetically Hardcore, Aggressive
No Mosaics. 无码 2倍速
INT8 Convrot for ComfyUI 0.27 [Native 原生节点支持]
File name: Krea2DarkBeast1.1-INT8-Convrot-ComfyUI (native nodes)
Reveal Every Fucking Detail - KREA 2 赤佬 Dark Beast Edition🐱👤06/28/2026
This is the raw, unfiltered experience you've been waiting for. KREA 2 赤佬 Dark Beast Edition strips away every layer of artificial restraint to deliver a brutally honest, detail-rich immersion. No fucking censorship, no blurred lines, no compromises—just pure, unadulterated content in its most authentic form.
We've gone beyond mere visual fidelity to capture the gritty textures, subtle imperfections, and visceral realism that define true hardcore experiences. Every pixel, every shadow, every nuance is rendered with obsessive precision, creating a world that feels alive, tangible, and dangerously real.
This isn't just another edition—it's a declaration. A middle finger to sanitized content and watered-down experiences. KREA 2 赤佬 Dark Beast Edition is for those who demand authenticity, who crave the unfiltered truth, and who refuse to settle for anything less than the complete, uncensored reality.
Usage ：ER_SDE/Euler | Simple | CFG=1 | 8-16Steps
链接：RedCraft | 红潮 | KREA 2 赤佬 Bastard Edition （Light Version）
INT8 Convrot for ComfyUI [Native] Uploaded
File name: Krea2RedMix1.1-INT8-Convrot-ComfyUI (original native)
File name: Krea2RedMix2.1-INT8-Convrot-ComfyUI (no mosaics)
Dark Beast 9 ZIT🟥DBZ in Moody Remixed for RClaw🦞 03/21/2026
Fast & Reliable Edition
This version is specially tuned for my🦞ClawBOT to deliver ultra-quick responses with significantly lower error rates. We fused the explosive energy and raw power of DBZ style into a battle-tested, community-matured open collaboration backbone, Moody Porn Mix 8 as the core base model. faster inference, sharper coherence, fewer hallucinations, and that signature intense, unapologetic vibe—perfect for users who want speed without sacrificing quality or personality. ClawBOT just got a Super Saiyan upgrade.
Recommended Settings: CFG Scale 1 Steps 8 Sampler & Scheduler Euler + Simple
THX
Moody Porn Mix
Poses, faces, details — everything you can think of has improved significantly. I honestly don’t know why I keep doing this to myself 😂 I keep saying, “This will be the final version,” but this model just can’t stop getting better 😏 @catlover1937
https://civarchive.com/models/620406
Dark Beast SDA 8 ZIT🟥Restoring Generative Diversity NSFW 03/15/2026
通过引用 @FOK 提出的新型 语义方向对齐（Semantic Directional Alignment, SDA）损失，本模型 merge 了 饼儿佬 训练的 LoKr 适配器，在完美保持 8 步基线模型极致锐度和快速推理速度的同时，恢复了 原始教师模型 70.2% 的组合多样性（LPIPS）
Designed to rescue the "Diversity Collapse" problem in few-step distilled Flow Matching / Diffusion models. By applying @FOK novel Semantic Directional Alignment (SDA) Loss, this Model adapter recovers 70.2% of the original teacher model's compositional diversity (LPIPS)
https://huggingface.co/F16/z-image-turbo-sda
DBZ 8 SDA Exported LoRa 版本模型 Rank256 FP16：
https://civarchive.com/models/964312?modelVersionId=2775165
DBZ 8 SDA workflow 配套工作流链接：
https://civarchive.com/models/579280?modelVersionId=2742099
✨经网友测试，V8 SDA 与 ZIT LoRA 存在兼容性问题（解锁自由度后无法对齐），正在与SDA作者沟通解决方案。
✨After testing by netizens, there is a compatibility issue between V8 SDA and ZIT LoRA (unable to align after unlocking degrees of freedom), and we are currently communicating with the SDA author for a solution.
Additional voice:
When distilling a 50-step diffusion model (teacher) into an 8-step model (student), the student tends to take shortcuts across the manifold and collapse toward the “mean” of the distribution. As a result, even when changing the random noise seed, the generated images are almost identical (with consistent character poses and layouts). Instead of directly matching the raw velocities, we fine-tune this model using a custom four-pillar architecture grounded in physical principles.
当将一个 50 步扩散模型（教师）蒸馏为 8 步模型（学生）时，学生模型往往会走捷径穿越流形，收敛到分布的“均值”。结果是，即使更改随机噪声种子，生成的图像也几乎完全相同（人物姿势、布局一致）。SDA方案没有直接匹配原始速度，而是使用一种基于物理原理的四支柱自定义架构对该模型进行微调（通过向量导航，老司机 FOK 的这套方法就像给流扩散模型装了“方向导航+安全带+聪明开关+最后刹车”，让它既敢大胆改构图，又不牺牲照片级清晰度。）
Through vector navigation, this approach is like equipping the flow diffusion model with "directional GPS + safety belt + smart switch + final brake," allowing it to boldly reshape compositions while preserving photo-level sharpness.
古有 Lvmin Zhang 控制网络锁姿势，今有 FOK佬 四条支柱破蒸馏！开源之基！🙏 LORD of OPEN DIFFUSION
Recommended Settings:
CFG Scale: 1
Steps: 8
Sampler & Scheduler: Euler + Simple
Key Features & Optimizations:
Restoring Diversity in 8-Steps
NSFW Unlock with 10% DPO color style and sharpness.
新手司机慎用，该模型极易造成肾衰竭
Dark Beast KLEIN 9b 🟦 V2.0 BFS 03/03/2026
由于系统故障例图工作流无法正确提取, DBK BFS workflow 独立链接：
https://civarchive.com/models/579280?modelVersionId=2740453
Dark Beast KLEIN 9b BFS 🟦 V2.0 LoRA Edition
https://civarchive.com/models/964312/redcraft-dark-beast-zandklein-exported-lora-edition
This is the next-level face-swap specialized evolution of the Dark Beast lineage, built on the lightning-fast FLUX.2 Klein 9B accelerated model from Black Forest Labs.
Engineered with targeted optimizations for face swapping practices, it integrates BFS (Best Face Swap) technology to completely eliminate the rigid, unnatural look that plagued earlier face replacements — delivering seamless, lifelike integrations with preserved identity, expression, and lighting.
It also fully fixes the portrait reference issue from the previous DB BlitZ versions, ensuring right reference adherence every time.
Special thanks to the scheme provider: https://github.com/alisson-anjos for the powerful BFS foundation that powers this breakthrough.🟦
The author's In-site link: https://civarchive.com/user/NRDX
Important notes:
This version is exclusively designed around the Klein 9B accelerated edition — no base model exists.
Usage is identical to Black Forest Labs' official FLUX.2 Klein 9B accelerated release: ultra-low steps (e.g., 4-5), CFG=1 fixed, blazing inference speed on consumer hardware.
In one sentence: Dark Beast's ferocious soul meets BFS (Best Face Swap) technology — more natural, and truly unstoppable! 🟦
for more infomation about BFS (Best Face Swap) :
https://huggingface.co/Alissonerdx
Alternatively, it can be directly applied to the entire Klein 9b/Qwen Edit base and Fine-tune models, through LoRA Adapter parameter injection.
面部交换测试的源-人像来自Civitai用户图像和Moody Mix肖像风格。
Source faces for the face-swapping test originated from Civitai user images and Moody Mix portraits styles.
Congratulations! DreamBooth training support for the Z-Image model is now officially available in HuggingFace Diffusers. Big thanks to that stunning dev from Tel Aviv! 🔥
[linoyts`s HF repo.]（https://huggingface.co/linoyts/beyond-reality-z-image-diffusers）
learn more about Diffusers Training：DreamBooth training example for Z-Image
Dark Beast BlitZ 6 ZIT🟥 NSFW models Zoo：
Pruned Model fp32 (22.93 GB) <- ComfyUI F32 精度模型权重
Pruned Model BF16 (11.46 GB) <- ComfyUI BF16 精度模型权重
Pruned Model fp8 (6.75 GB) <- ComfyUI Scaled FP8 Mixed 混合精度
Pruned Model nf4 (6.73 GB) <- NVFP4 Mixed 混合精度（BLACKWELL 50系加速）
Training Data (3.75 KB) <- ComfyUI Simple Hybrid Workflow 简易混合采样工作流
Or download from the "Files" list below the "Details" on the right side of this page>>
Dark Beast BlitZ 6 ZIT🟥BeyondReality3 NSFW 02/18/2026
based on SUPER Z V3.0 REBUILD (Vibrant Color Edition)
NSFW Unlocked Version of Beyond Reality 3 form [@Nurburgring`s HF repo.](https://huggingface.co/Nurburgring/BEYOND_REALITY_Z_IMAGE)
核心改进：
提供卓越的色彩和照明性能，具有更精致的纹理细节。增强功能：与 ZiTurbo 相比，支持更广泛的艺术风格，同时提供比 ZI Base 更快的生成速度。NSFW已解锁
Core Improvements:
Delivers superior color and lighting performance, with more refined texture details. Enhanced Capabilities: Supports a wider range of artistic styles compared to Z image turbo, while offering faster generation speeds than Z image base. NSFW Unlocked
Recommended Settings:
CFG Scale: 1
Steps: 5～15
Sampler & Scheduler: Euler + Simple
Key Features & Optimizations:
NSFW Unlock and maintain BR3 color style and sharpness.
增加画面的多样性表达，优化了部分艺术风格支持与二次元动漫表现。
Optimized support for specific artistic styles and 2D Anime rendering.
[HF repo.](https://huggingface.co/GuangyuanSD/REDCraft-DarkBeast-Z-Image-TURBO)
Dark Beast BlitZ 6 ZIT🟥 NSFW LoRA Adapter 02/18/2026
Additionally, I've exported Dark Beast BlitZ 6 in Rank-256 LoRA format. The LoRA weight can be adjusted to adapt it to various ZIT fine-tune models, fully compatible with the Z-Image Turbo base model.
Full Model fp16 (1.06 GB) <- 可以通过这里直接下载 LoRA 版本
上面是 Dark Beast BlitZ 6 导出为 Rank256 的LoRA版本，可以调整权重强度用于各种微调ZIT版本
（适配于 ZimageTurbo 基底模型）
Dark Beast BlitZ 6 ZIT🟥 NSFW LoRA adaptation models Zoo：
Z-Image-Turbo-GGUF <- Z-Image Turbo & TE GGUF 量化模型
Z-Image Turbo - Nunchaku Quantized <- Z-Image Turbo & TE Nunchaku 量化模型
Text Encoder (ClipLoader use) <- Qwen3 4b FP16 文本编码器
Abliterated Huihui Qwen3 4B v2 (Q_8 GGUF) <- Z-Image Uncensored TE 文本编码器
VAE (Flux.1 16C VAE) <- 标准的 Flux.1 16 通道 VAE
Or download from the "Files" list below the "Details" on the right side of this page>>
Dark Beast KLEIN 9b BlitZ🟦V1.5 lora adapter 02/16/2026
DarkBeastKlein-v1-5-Blitz-9b-Extracted-LoRA-r256 uploaded
working fine with FLUX.2 Klein 9b models
Pruned Model fp16 (1.23 GB) <<< This one
Dark Beast BlitZ ZIT🟥5-steps lora adapter 02/16/2026
DarkBeast5steps_extracted_lora_r256 uploaded
working fine with Zimage turbo models
Pruned Model fp16 (1.04 GB) <<< This one
Or download it from the "Files" list below the "Details" on the right side of this page>>
Dark Beast BlitZ ZIT🟥5-steps 02/12/2026
一句话：以 ZImage Turbo 已经炽热的速度，
将其调到绝对 BlitZ 速度，现在就释放出来！
只需5步，同时保留那凶猛的黑暗野兽灵魂的每一滴！ 🟥
Key features:
Fully preserves the signature Dark Beast style, rich details, and intense Black Beast aesthetic from the standard lineage
Refined through advanced targeted distillation & fine-tuning, now perfectly dialed in for zero-CFG guidance at minimal steps
BlitZ-level inference speed — breathtaking high-quality images in just 5 steps
Recommended settings: 5 steps, CFG=1 (fixed), any seed you want
In one sentence: Taking ZImage Turbo’s already blazing speed and cranking it to absolute BlitZ velocity while keeping every drop of that ferocious Dark Beast soul! 🟥
Lightning-fast generation awaits — unleash it now! 🟥🚀
Dark Beast KLEIN 9b🟦V1.5 BlitZ 02/08/2026
This is the ultimate speed-optimized V1 evolution, powered by Flux.2 Klein 9B,
engineered specifically for lightning-fast low-step + CFG=1 workflows (5steps).
Also available in NVFP4 quantized format, optimized for acceleration on Blackwell architecture GPUs.
（ like RTX50XX, PRO6000, B200, and others ）
Also supports non-50 series GPUs (automatic 16-bit operation), Verify environment is my ComfyUI 0.11
Key features:
Fully preserves the signature Dark Beast style, rich details, and intense Black Beast
aesthetic from the standard lineageRefined through advanced targeted distillation & fine-tuning, now perfectly dialed
in for zero-CFG guidance at minimal stepsBlitZ-level inference speed — breathtaking high-quality images in just 5 steps ⚡
Recommended settings: 5 steps, CFG=1 (fixed), any seed you want
In one sentence: Taking Klein’s already blazing speed and cranking it to absolute BlitZ
velocity while keeping every drop of that ferocious Dark Beast soul! 🟦
Lightning-fast generation awaits — unleash it now! 🚀
Basic workflow: Small-scale composition + large-scale subsampling,
followed by final redrawing (SeedVR2).
The basic mixing scheme for DarkBeast KLEIN 9b V1:
This is a speed-optimized version specially tuned for Flux.2 Klein 9B base,
designed for the low-step + CFG=1-2 scenario (10–20 steps).
Key features:
Inherits the core style and detail quality from the Dark Beast standard version
Through targeted distillation & fine-tuning, perfectly adapted for low steps + no CFG guidance
Extremely fast inference (high-quality images in just 10–20 steps)
Recommended inference settings: 10–20 steps, CFG=1 (fixed), seed whatever you like
In one sentence: Pushing Klein's speed to the absolute limit while still keeping that signature Black Beast flavor! 🟦
Super fast generation — give it a try! ⚡
Big thanks to @alcaitiff for the awesome work and killer contributions to training Z-Image Turbo and Klein models! Seriously impressive stuff! 🚀
非常感谢 @alcaitiff 对 Zimage turbo 和 Klein 9b 的模型训练做出的杰出贡献！
https://civarchive.com/user/alcaitiff
https://civarchive.com/models/2348977/klein-9b-unchained-xxx
https://civarchive.com/models/2206377/zit-mystic-xxx
Dark Beast Z-造相·黑兽-DBZ⚡️Final
迎接 Zimage non-turbo 版本 ！特此放出基于ZiT的“黑兽”最终版本，期待官方EDIT版本的发布。我们会继续在ZI的基础上进行强化学习，并继续以 Turbo 方式发布全新 sft models
修正黑兽I-III越来越过激越界的调试方案，回归本质。
DBZ IV 是基于RedZ1.5 重新调试的修正畸变版本（没有额外突破性尝试）
构图能力对标 ZiT base 原版，所有概念元素齐全，LoRA适配性强。
---
Introducing Zimage DBZ IV sft FINAL model
Today we proudly release the final version of "Dark Beast ZiT" (黑兽),
fully rebuilt and refined on the ZiT foundation.
This iteration addresses and corrects the increasingly extreme and boundary-pushing tuning drifts seen in Dark Beast I–III, returning firmly to its core essence and original intent.
DBZ IV is a carefully re-tuned and distortion-corrected variant built directly on RedZ1.5.
It introduces no groundbreaking new capabilities beyond solid refinements
—its strengths lie in stability and fidelity.
Composition quality matches or closely rivals the original ZiT base model.
Full coverage of conceptual elements and prompt adherence.
Excellent LoRA compatibility and strong adapter tolerance across a wide range of styles and concepts.
Enjoy the refined Dark Beast experience—clean, powerful, and true to its roots.
Feel free to tweak this further if you want a more hype-driven tone, shorter blurb,
or additional technical details!
推荐生成设置
分辨率：优化适用于 1M~2M 像素（如 1048-1536 long-edge）
CFG Scale：1
Sampler：Euler + SGM Uniform 或 ER SDE
Steps：8–10（内置 Ultra Flux VAE）
Recommended Generation Settings
Resolution: Optimized for 1M~2M pixels (1048-1536 long-edge)
CFG Scale: 1
Sampler: Euler + SGM Uniform or ER SDE
Steps: 8–10 (Ultra Flux VAE inclouded)
THX ！@Flux_dev for open-source high-quality Northeast Asian female character LoRAs
Author of the prompts in the above picture (source) @iamddtla @Cosmic_Crafter
DBZ⚡️image4K 黑兽- UltraSHAPE v31 01/16/26
Dark Beast Z 3 Ultra SHAPE 一款专为顶级创作者打造的融合模型，旨在实现超越现实边界的极致超写实情色影像。通过Dype scale与SeedVR2，该模型可生成令人惊艳的 16M 像素输出。
Recommended to learn how to configure DyPe @FloyoAI:
DyPe and Z-Image Turbo for High Quality Text to Image
https://civarchive.com/articles/24570/dype-and-z-image-turbo-for-high-quality-text-to-image
推荐以下文章了解如何配置 DyPe High Quality 4K T2I
重要说明
关于DBZ3.1 经群友测试发现一些特定的质量词会导致整体偏移出错，我的解决方办法有：
1 删除一部分非必要的质量描述
Remove some non-essential quality descriptions
2 DyPe 设置中选定方法为 Zimage，以及尝试 ntk 或者 base 方式
In the DyPe settings, the selected method is Zimage, and we also tried ntk or base methods
3 将DyPe 设置中的 max_shift 上限开大
Increase the upper limit of max_shift in the DyPe settings
推荐生成设置
分辨率：优化适用于 4M~16M 像素（如 2048-4096 long-edge）
CFG Scale：1
Sampler：Euler + SGM Uniform 或 ER SDE
Steps：8–10（推荐 Ultra Flux VAE）
Recommended Generation Settings
Resolution: Optimized for 4M–16M pixels (2048-4096 long-edge)
CFG Scale: 1
Sampler: Euler + SGM Uniform or ER SDE
Steps: 8–10 (Ultra Flux VAE recommended)
DBZ⚡️image 红潮黑兽 v25 1/8/26
DBZII 黑兽2.5 重新找回Tags无情抽卡时的童年乐趣
DBZ 2 : rediscovering that pure, unfiltered gacha dopamine rush
when you finally pull the perfect tags
after 100 cursed rolls 💗
---
这是「LSP」Lovin' Snuggly Positions v1的 Turbo base 版本
推荐生成设置
分辨率：优化适用于 1.2M~2M 像素（如 1536x832 1920×1080 或类似）
CFG Scale：1（强泛化能力适用于Tags抽卡，保持对提示的自然响应）
Sampler：Euler + SGM Uniform 或 ER SDE
Steps：8–10（不推荐 Ultra Flux VAE）
Recommended Generation Settings
Resolution: Optimized for 1.2M–2M pixels (e.g., 1536×832, 1920×1080)
CFG Scale: 1 (Strong generalization for tags prompts, preserves natural prompt adherence)
Sampler: Euler + SGM Uniform or ER SDE
Steps: 8–10 (Ultra Flux VAE not recommended)
Comparison of Same Prompts and Seeds：
DBZ⚡️image 红潮黑兽 AIO 12/21/25
based on BEYOND REALITY 超越真实 The author is [受保密条款保护]
https://civarchive.com/models/1090420/beyond-reality
Dark Beast Z Super Fusion – 模型介绍
Dark Beast Z Super Fusion 是一款专为顶级创作者打造的融合模型，旨在实现超越现实边界的极致超写实情色影像。通过精妙融合 MysticXXX ZIT v2 的强烈诱惑写实感、Kawaii Kinky[NE] v1 的大胆挑逗能量，以及 Detail DeamonZ 无与伦比的精细细节掌控能力，并以突破性的超摄影写实底模 BEYOND REALITY sft ZiTurbo 为基础，该模型可生成令人惊艳的 2M 像素输出，具备完美无瑕的皮肤纹理、栩栩如生的面部表情，以及超越真实摄影的超精细光影表现——即使在最复杂且暗示性的构图中亦游刃有余。
最终成果是一个极具多功能性的检查点，擅长渲染超越现实的视觉效果：想象皮肤渲染得如此精湛，能清晰呈现每一个毛孔与细微血管；眼神深邃仿佛能直刺观者灵魂；身体兼具解剖学完美与自然瑕疵，搭配挑逗姿势、精致内衣、复杂束缚元素，以及传达纯粹、未经滤镜的欲望与强烈情感的表情。
融合流程概述
融合过程采用 ComfyUI 工作流中的高级块加权合并技术，精准执行：
底模建立：以 BEYOND REALITY sft ZiTurbo 作为核心模型（Model A），充分利用其无与伦比的摄影级写实基础与卓越解剖准确性，为超写实奠定基础。
主要融合：以针对性权重（关键块如 base、inpaint、VAEs 为 0.6–0.8）融入 MysticXXX ZIT v2，注入深度高保真情色写实感，并强化皮肤渲染与动态光影表现。
风格注入：在中上层块选择性合并 Kawaii Kinky[NE] v1（强度 0.6–0.8），引入挑逗面部动态、性感肢体语言与情色张力，同时保持整体写实主导。
细节精炼：最后以较低权重（-1.0）应用 Detail DeamonZ，专注纹理增强与精细细节块，实现眼睛、发丝、织物微纹、汗珠及皮肤细微瑕疵的极致锐利清晰，推动输出进入“超越现实”领域。
全程进行多轮迭代测试，确保无伪影并完美平衡原始情色与超越摄影的写实美感。
推荐生成设置
分辨率：优化适用于 2M 像素（如 1920×1080 或类似）
CFG Scale：1–2（保持对提示的自然响应）
Sampler：Euler + SGM Uniform 或 ER SDE
Steps：8–10（高效超高质量输出）
Clip Skip：1–2
专为追求超越摄影极限的极致超写实成人影像的艺术家而生，能够捕捉欲望与细节的每一丝细微之处。
The DBZ1.0 training set comes from the ultimate essence of the community
⚡️20+ year veterans' lightning-fast meme-battle handspeed.(Thanks to [受保密条款保护]，佬杨⚡️@wikeeyang, 亮总⚡️@亮亮rayne, 十字鱼⚡️@Gluttony100,可乐⚡️@Colour, and everyone else⚡️⚡️⚡️⚡️⚡️⚡️)
Dark Beast Z Super Fusion – Model Introduction
Dark Beast Z Super Fusion is a fusion model crafted exclusively for top-tier creators, designed to achieve ultimate hyper-realistic erotic imagery that transcends the boundaries of reality. Through masterful blending of the intense seductive realism from MysticXXX ZIT v2, the bold provocative energy of Kawaii Kinky[NE] 小恶魔 v1, and the unparalleled fine-detail mastery of Detail DeamonZ , all built upon the groundbreaking ultra-photorealistic base of BEYOND REALITY 超越真实 sft ZiTurbo, this model generates breathtaking 2M-pixel outputs with flawless skin textures, lifelike facial expressions, and hyper-detailed lighting that surpasses real-world photography—even in the most complex and suggestive compositions.
The result is an extraordinarily versatile checkpoint that excels at rendering visuals of transcendent realism: imagine skin rendered with such precision that every pore and subtle vein is clearly visible; eyes with profound depth that seem to pierce the viewer's soul; bodies combining anatomical perfection with natural imperfections, paired with provocative poses, exquisite lingerie, intricate bondage elements, and expressions conveying pure, unfiltered desire and intense emotion.
Fusion Process Overview
The fusion was executed with precision using advanced block-weighted merging techniques in ComfyUI workflows:
Base Establishment: Started with BEYOND REALITY sft ZiTurbo as the core model (Model A), leveraging its unmatched photographic-level realism and superior anatomical accuracy to lay the foundation for hyper-realism.
Primary Integration: Incorporated MysticXXX ZIT v2 at targeted weights (0.6–0.8 across key blocks such as base, inpaint, and VAEs) to infuse deep, high-fidelity erotic realism while enhancing skin rendering and dynamic lighting behavior.
Style Infusion: Selectively merged Kawaii Kinky[NE] v1 in mid-to-upper layers (intensity 0.6–0.8) to introduce provocative facial dynamics, sensual body language, and erotic tension while maintaining overall realism dominance.
Detail Refinement: Applied a final pass with Detail DeamonZ at lower weights (-1.0), focused on texture-enhancing and fine-detail blocks to achieve razor-sharp clarity in eyes, hair strands, fabric micro-textures, sweat beads, and subtle skin imperfections—pushing the output into "beyond reality" territory.
Multiple iterative tests were conducted throughout to ensure artifact-free results and perfect balance between raw eroticism and transcendent photographic realism.
Recommended Generation Settings
Resolution: Optimized for 2M pixels (e.g., 1920×1080 or similar)
CFG Scale: 1–2 (maintains natural adherence to prompts)
Sampler: Euler + SGM Uniform or ER SDE
Steps: 8–10 (efficient, ultra-high-quality output)
Clip Skip: 1–2
Created for artists pursuing the pinnacle of hyper-realistic adult imagery that exceeds photographic limits, capable of capturing every nuance of desire and detail.
冬至日快乐
Forever in memory of METAFILM Studio founder Mr. Yuan Bo
Description
Dark Beast 6 BlitZ ZIT based on SUPER Z V3.0 REBUILD (Vibrant Color Edition)
NSFW Unlocked Version: Beyond Reality 3 form [@Nurburgring`s HF repo](https://huggingface.co/Nurburgring/BEYOND_REALITY_Z_IMAGE)
Core Improvements:
Delivers superior color and lighting performance, with more refined texture details. Enhanced Capabilities: Supports a wider range of artistic styles compared to Z image turbo, while offering faster generation speeds than Z image base. NSFW Unlocked
Recommended Settings:
CFG Scale: 1
Steps: 5～15
Sampler & Scheduler: Euler + Simple
Key Features & Optimizations:
NSFW Unlock and maintain BR3 color style and sharpness.
Optimized support for specific artistic styles and 2D Anime rendering.
[HF repo](https://huggingface.co/GuangyuanSD/REDCraft-DarkBeast-Z-Image-TURBO)
FAQ
Comments (79)
Thanks for your hard work, everyone! Thank you for updating during the holiday; there's new stuff to play with now.
大佬們辛苦了～感謝年假期間更新，有新東西玩了
Thankyou for the beautiful model! But why can't my character Lora (trained with AI toolkit) works with this model?
LoRA compatibility for DBZ6 is on par with other ZIT fine-tunes — it works seamlessly across most variants.
For the optimal adaptation and highest fidelity (best concept capture, least drift), pair it with the dedicated training adapter designed for Z-Image Turbo/ZIT workflows (e.g., from Ostris AI Toolkit or similar). This de-distills during training/inference for cleaner, more accurate results.
@AiMetatron thankyou for your guidance 👏🏻
I tested BeyondNSFW, both FP8 and FP32. Both are absolutely fantastic. If you have the option to use the FP32, it's slightly better overall, but the FP8 also produces near-perfect images in my opinion. Ultra-realistic, vibrant images that follow the prompts very good. Unfortunately, the text handling isn't very good, that's the only problem I see. I'll be using BF16 myself, i just like it the best. Thank you for this work, i love it.
What can you say about the nf4 model? Is it as good as fp8 or fp32?
@lolopopo123 I haven't tested NF4. However, I suspect the quality will be very high, just like the other three.
YEA, BF16 is the most balanced and stable quantization specification. F32 will improve the precision of details, and FP8Mixed will occupy less CPU and GPU memory.
@lolopopo123 NVFP4 quality ≈ FP8, but 2x faster on 50-series GPUs. Works elsewhere too, though VRAM usage ≈ BF16 (worse than FP8-Mixed).
If you're not sure what spec/version to pick, just grab the LoRA version of DBZ6. Pair it with whatever ZIT model you've already got locally — super easy to use and tweak =）
@AiMetatron I have a 5060ti 16gb, is the NVFP4 faster without losing quality than the fp8? And thank you for making the models, I'm currently using the DBZ ZIT and I'm having a great time with it!
@lolopopo123 I have a 4070 with 12GB, with LLM, i'd rather wait 30 seconds for the quality to improve than have a worse picture in 20 seconds. I'd even wait 50 seconds. It's other then with SDXL/ILLU/PONY... my opinion
@Gorean How long does it take to run the BF16 model in your machine? I often oomed or crash when I use a fp16 model that's why I stay away from bigger inferences but kinda curious now because of your experience. Is it alright for you to share your workflow? Thank you very much!
@lolopopo123 i use two workflow, the simple one from comfy and one with 3 ksampler. I don't use the portable version, i use the desktop version. With the simple workflow i get in 6 Steps a nice picture with 1440x1920 in ca 30 sec, with the 3 ksampler it needs ca 43 sec and is mostly much better.
@AiMetatron Is the LORA version the 1gb one: darkblitz6Extracted.14OH.safetensors?
When I try NSFW, the generated images always show anal sex, even if I explicitly state that it must be inserted into the vagina. The images are always inserted into the anus, causing me great pain and seeking help
@lolopopo123 Since you already have 16GB of video memory, it is recommended to use BF16 for better performance. NVFP4 is mainly suitable for 50 series GPUs under 8GB. We expect NVFP16 to achieve maximum quality and speed, but unfortunately, there is currently no specification sheet available for reference.
@LazarusTricks right
Will fp16 and fp8 versions of DBK be available?
Z6的上限似乎更高，但整体来讲出现结构问题的概率远大于Z5，画面中的杂乱元素野更多，导致成功率变低不少，希望能优化。
主要是锐化程度和鲜艳程度更高了（得益于BR3的调教）结构稳定和杂质的话，相比DBZ早期过激的尝试还是要好一些的，跟Z5比较起来更讨好眼球。对于最大的稳定性和技术突破，还是会放在红潮系列来进行更新。
红潮2其实早就训练好了...但是一直不允许放出，官方训练器好像也没有出来...所以红潮版本，只好在ZIB和其他方面再做尝试。
@AiMetatron 了解了。我之前用Zimage turbo和Z5的时候都是会用两个高级ksampler跳过中间采样步数，保留最开始步骤的提示遵循能力，再用靠后的步数直接优化细节。而Z6的话难以采取这个工作流了，Z6在后面的采样步数里不太对结构产生影响了，而在中间步数的影响又过大，似乎整个采样流程比较生硬地从大结构直接跳转到小结构了。只用一个采样器来简单生成的话，还是存在质量和提示遵循难兼得的情况。Civitai现在会禁止模型公布吗，是个什么情况
@AiMetatron 你好大佬，想问下这个黑兽系列和红潮的区别是什么啊？比如在NSFW内容上生成的东西有什么不同
It's an interesting model. I haven't tested it much yet, but I can already say that it does an excellent job with my tasks. Compared to the basic ZIT, your model is brighter, more colourful, and, in my opinion, interacts better with Lora. I would also like to note that there are fewer limb issues than with other models. It does happen sometimes, but it's nothing serious. Thank you, I think I will replace the basic model with yours. Thank you very much for sharing it. I am using version bf16.
Thank you so much for your testing and valuable feedback! 🙏
DBZ6 is actually a milestone version for the Zimage model series. It incorporates comprehensive calibration across various aspects, and we specially released it during this Lunar New Year season for everyone to try out.
In addition, I have also uploaded the DBZ6 LoRA versions on this page. These are designed to be compatible with various platforms and quantization environments, such as GGUF, Nunchaku, MLX, DF11, and ONNX.
Full Model fp16 (1.06 GB) << seems to be truncated/not uploaded 完整模型 fp16
Nope, this is an exported DBZ6 LoRA version Rank256 =）
@AiMetatron sorry I need some more clarification. Do I download the the 11 Gig Model and Run the 1.06 Gig as a LoRA, or do I run the 1.06 gig as a Diffusion Model. If it's the latter I get a UNET Error. Is there something special I need to do in Comfyui? (I copied your Nodes Directly from an Image from the 1.06gig version for testing).
@Neeco Dark Beast BlitZ 6 ZIT🟥 NSFW LoRA Adapter 02/18/2026
Additionally, I've exported Dark Beast BlitZ 6 in Rank-256 LoRA format. The LoRA weight can be adjusted to adapt it to various ZIT fine-tune models（or Quantitative format like GGUF Nunchaku MLX, fully compatible with the Z-Image Turbo base model.
@AiMetatron Ah okay, thanks I guess I missed that in your original post, thanks for the response. :) @oceanswave sorry for hijacking your comment. ;)
这个要放在哪个文件夹？求赐教
Dark Beast Blitz 6 LoRa is awesome! It has always been a problem that unofficial checkpoints change the face of my character LoRas. That's over! Dark Beast BlitZ 6 LoRa with 0.65-0.7 strength + character LoRa with 0.9-1.0 strength is perfect! I recommend using the Selective LoRa Loader for the character LoRas.
Hi, could you please tell me what "Selective LoRa Loader" is? Is it a node in COMFYUI?
@wyxzddsjj919 Yes, it's a custom node. You can download via the UIManager, or from here. GitHub - shootthesound/comfyUI-Realtime-Lora: Train and block edit and save LoRAs directly inside ComfyUI for Z-image, Flux Klein, SDXL, Flux, WAN 2.2, SD 1.5
@burcsi271 Thank you for your reply.❤️
What settings would you suggest for the analisys loader?
@ashdodd1029515 Not the analyzer, use the Selective LoRA Loader (Z-Image) with Face Priority Agressive preset.
@burcsi271 And you run the other loras at default? Like a pose lora and a face lora for example. Running face lora in aggressive, you run the pose as default? Or do you oppose the face lora's layer weights?
@ashdodd1029515 I'm not using a pose lora. Dark Beast BlitZ 6 LoRa with 0.65-0.7 strength + character LoRa with 0.9-1.0 strength. Pose and everything else is prompt.
The quality is nice but seems to be bleeding a bit too heavily, overriding the styles/characters from model, even when using lower strenghts like 0.2-0.4.
Thank you for your feedback
When I try NSFW, the generated images always show anal sex, even if I explicitly state that it must be inserted into the vagina. The images are always inserted into the anus, causing me great pain and seeking help
This is a long-standing issue that has existed since the XL and PONY eras, depending on the size of the composition and the description of the posture.
Hello - I am trying to use your Dark Beast KLEIN 9b🟦V1.5 BlitZ in ComfyUI, but the file you have attached in Civitai will not execute. I get the following error: ERROR: Could not detect model type of: /workspace/runpod-slim/ComfyUI/models/diffusion_models/Flux2Klein_9B/darkBeastFeb1826Latest_dbkBlitzV15.safetensors
I successfully downloaded it via the Civicomfy extension into my diffusion_models folder, but I think there may be something wrong with the file itself.
Can you confirm the type of model you downloaded? Because each version contains a large number of quantization formats (or independent Lora exports) in ComfyUI, users need various specifications, so we upload as many files as possible in the file list on the right side of the page. Don't just trust the official recommended blue button (choose the version you need based on the model introduction page and the files list on the right)
@AiMetatron Thanks for the reply! I did not realize there were 6 files under the KLEIN 9b🟦V1.5 BlitZ model. I was pretty certain that Civicomfy was pulling in the darkBeastFeb1826Latest_dbkBlitzV15.safetensors file. But I will try downloading the file and importing it straight into ComfyUI to see if I get a different outcome.
bf16...Definitely not a bad model, honestly I will have to make a new workflow for this ,also Euler + Simple give worse results for me, I also have to adjust the Lora models, all in all not bad! Thanks for sharing...what about fp16
FP16 is linked to the Lora version of this model, so the file size is relatively small
just a second bro, I don't understand ,so fp16 from 1GB is actually Lora not chekpoint, but it is on the list with chekpoints ! If it is Lora, can it be used with these checkpoints in combination and if so, does it have any additional effect that enables on bf16
为什么人物皮肤周围会出现马赛克一样的线条，感觉是要表现出人体的汗毛最后变成一整条线了
能具体说说吗？euler simple 采样吗？
@AiMetatron 对，就是用你示例图的工作流，有概率出现这种情况，尤其是背景光照在身上，人物手臂上的绒毛就会糊成一条条马赛克一样的线
LoRA加上后，人物的脸都变形了，不知道是什么原因
Really like your lora for klein9 !
Thank you
任何模型都不如原版好用，这是不争的事实
也包括XL和F1吗？试用了Klein吗？PONY7呢？Z确实官方做了很成功的Turbo，这个是不争的，但是如果因此得出Finne-tune无用论，真的以偏概全了
@AiMetatron 这个模型不错，只是加上LoRA后就产生奇怪的画面，而且人物变形很怪异，大佬有没有什么好办法可以教教小白啊？望回复。
a little confused on what is a Lora and what is a checkpoint, its labeled as checkpoint, but there are a dozen different models with huge variation in size
The small 1gb file is the lora.
CAUTION : HIGH VALUES RESULTING IN INSANE QUALITY.
CFG:2,STEPS:15, DETAILER DEMAON AT .60, TOP TIER LEVEL IMAGES( at least for me, since I love photorealistic stuff)
May I kindly ask if you could share your WF and what hardware you have and gen times?
@SaburoDio He is generating online with no hardware requirements...
@SaburoDio Just a normal WF, since I'm generating online as @XpPillow said, so it is much easier for me to generate images without stressing my laptop hehe
@6vidit9 So you are into z-image too lol
@XpPillow I'm into everything XD
@6vidit9 thanks, I was hoping for a magical workflow that I could run on my 4060 haha
@SaburoDio 4060 is capable to run it as long as you have 64gb+ ram
@SaburoDio I'm sure you can, 4060 is pretty amazing GPU, if you'd look around here, community has lot to offer, my best advice for you to look for their discord server ask them, they will def gonna help you :)
@XpPillow 64gb 😭
@XpPillow and @6vidit9 and @XpPillow
Thanks for your replies! I have a feeling that this community is really supportive.
I got 16 GB of RAM XD
I am able to use ZIT checkpoints up to 12GB in size. I am mostly really happy with the outputs and time it takes but reaching "Tier" level is still a mistery to me.
Right now I am using the simple ZIT wf by subtleshader. I could not get the SeedVr workflow from one of the other creators to run properly as I got OOMs.
The discord is a great tip and I will try to find some anwers there :) Appreciate ya'll
Dark Beast BlitZ 6 ZIT REVIEW:
2-20-2026
-I have tested this checkpoint so far by copy/paste the shared photos on the page and I just have to say...dang it! It's so good. It works right out of the box. The photos with no make-up show realistic skin and the ones that show perfect skin shows perfect skin.
-The vibe/quality is as good as the photos that z-image developers showed us on the hugging face page. The output images are bright and vibrant. I'm using Forge Neo, so I don't get the seedvr2 to upscale and add the perfect skin, but the upside is that I get that quality, crisp, detailed RAW photos on some of the generations. Seedvr2 is not needed though as with the correct prompt the perfect skin can be obtained.
Amazing checkpoint, really like it!! thanks
thx
Details
Files
darkBeast5CSEP0BUZZH3_darkblitz6Beyondnsfw_pruned_bf16.safetensors
Mirrors
dark_beast_z6.safetensors
DarkBeastZ6-BlitZ-BF16-ComfyUI.safetensors
DarkBeastZ6-BlitZ-BF16-ComfyUI.safetensors
darkBeastFeb2226Latest_darkblitz6Beyondnsfw.safetensors
dark_beast_z6.safetensors
dark_beast_z6.safetensors
dark-beast-v15mar.safetensors
DarkBeastZ6-BlitZ-BF16-ComfyUI.safetensors
darkblitz6Beyondnsfw_bf16.safetensors
dark_beast_darkblitz6_bf16_v6_by_aimetatron.safetensors
darkBeast_darkblitz6Beyondnsfw.safetensors
darkBeastFeb2226Latest_darkblitz6Beyondnsfw.safetensors
darkBeast5CSEP0BUZZH3_darkblitz6Beyondnsfw_full_fp16.safetensors
Mirrors
darkBeast5CSEP0BUZZH3_darkblitz6Beyondnsfw_pruned_fp8.safetensors
Mirrors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.