文字无拘，视觉无界，艺术无限。Words Unbound, Vision Unleashed, Art Unlimited.RH体验地址（邀请码Invite Code: rh-v1603）<a target="_blank" rel="ugc" href="https://www.runninghub.ai/user-center/1874327105726242818/webapp?inviteCode=rh-v1603">https://www.runninghub.ai/user-center/1874327105726242818/webapp?inviteCode=rh-v1603</a>通过邀请链接注册领1000RH币，每天登录还可以领取100RH币，可以免费畅快跑图了！Sign up through the invite link to get 1,000 RH coins! Plus, log in daily to claim 100 more RH coins. You can now generate images for free!My Youtube：<a target="_blank" rel="ugc" href="https://www.youtube.com/@mugicha-AI">https://www.youtube.com/@mugicha-AI</a><span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">如非作者同意禁止搬运。Please do not redistribute without consent.<hr />2026-8-08该版本对数据序列进行了重新校准，在进一步强化人物真实感的同时，显著提升了画面美学质感与光影层次。相较于 SinoXvision‑krea2‑Uncen，该版本与原版 Turbo 的风格差异被进一步拉大。在线试用地址：<a target="_blank" rel="ugc" href="https://www.runninghub.ai/ai-detail/2087010178864984066?inviteCode=rh-v1603">https://www.runninghub.ai/ai-detail/2087010178864984066?inviteCode=rh-v1603</a>使用工作流：<a target="_blank" rel="ugc" href="https://www.runninghub.ai/zh-cn/post/2074520959322779649/?inviteCode=rh-v1603">https://www.runninghub.ai/zh-cn/post/2074520959322779649/?inviteCode=rh-v1603</a>This version recalibrates the data sequence, further enhancing character realism while significantly elevating the visual aesthetics and lighting depth. Compared to SinoXvision‑krea2‑Uncen, the stylistic gap between this version and the original Turbo has been widened even further.Online trial address：<a target="_blank" rel="ugc" href="https://www.runninghub.ai/ai-detail/2087010178864984066?inviteCode=rh-v1603">https://www.runninghub.ai/ai-detail/2087010178864984066?inviteCode=rh-v1603</a><a target="_blank" rel="ugc" href="https://www.runninghub.ai/zh-cn/post/2074520959322779649/?inviteCode=rh-v1603">Use workflow：https://www.runninghub.ai/zh-cn/post/2074520959322779649/?inviteCode=rh-v1603</a><hr />2026-7-29中文介绍 / Chinese Description<span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">Edit_v1.1 版本集成最新 krea2-identity-edit V1.2，图像编辑准确性与指令遵循度大幅提升；同时针对人像摄影美学进行专项强化，进一步消除数码感，使皮肤肌理、发丝细节、光影层次、衣物材质及远景景深过渡更贴近高端光学镜头质感，真实度与美学表现力显著增强，并继续保持破限支持，为创意提供更自由、更广阔的发挥空间。<span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">在线试用地址：<a target="_blank" rel="ugc" href="https://www.runninghub.ai/ai-detail/2083545404389462017?inviteCode=rh-v1603"><span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">https://www.runninghub.ai/ai-detail/2083545404389462017?inviteCode=rh-v1603</a>请搭配krea2图像编辑多功能工作流使用：<a target="_blank" rel="ugc" href="https://www.runninghub.ai/post/2078038950391943170/?inviteCode=rh-v1603">https://www.runninghub.ai/post/2078038950391943170/?inviteCode=rh-v1603</a>英文介绍 / English Description<span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">Edit_v1.1 integrates the latest krea2-identity-edit V1.2, delivering significant improvements in image editing accuracy and instruction adherence. It also features targeted enhancements in portrait photography aesthetics, further eliminating digital artifacts to bring skin texture, hair details, light-shadow transitions, fabric materials, and background depth-of-field closer to the quality of high-end optical lenses, with notably improved realism and aesthetic appeal. Breakthrough support remains enabled, offering greater creative freedom and broader possibilities.model should be used with the Krea2 multi function image editing workflow：Online trial address：<a target="_blank" rel="ugc" href="https://www.runninghub.ai/ai-detail/2083545404389462017?inviteCode=rh-v1603"><span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">https://www.runninghub.ai/ai-detail/2083545404389462017?inviteCode=rh-v1603</a><a target="_blank" rel="ugc" href="https://www.runninghub.ai/post/2078038950391943170/?inviteCode=rh-v1603">https://www.runninghub.ai/post/2078038950391943170/?inviteCode=rh-v1603</a><hr />2026-7-17中文介绍 / Chinese DescriptionEdit_v1.0模型请搭配krea2图像编辑多功能工作流使用：<a target="_blank" rel="ugc" href="https://www.runninghub.ai/post/2078038950391943170/?inviteCode=rh-v1603">https://www.runninghub.ai/post/2078038950391943170/?inviteCode=rh-v1603</a>Edit_v1.0基于 Krea2 Turbo 深度微调，在完美保留极速推理与原生美学的前提下，定向消除了人像“塑料感”，使皮肤发丝、光影过渡、衣物配饰及远景细节更具真实摄影质感，支持破限。并融入了 krea2-identity-edit 图像编辑能力，原生支持「文生图」、「局部图像编辑」与「图像特征参考」，实现极速、全能创作。提示： 尽管本模型具备图像编辑能力，但其专注度和精细度目前尚不及 Qwen-Image-Edit 等原生垂直模型英文介绍 / English DescriptionThe Edit_v1.0 model should be used with the Krea2 multi function image editing workflow：<a target="_blank" rel="ugc" href="https://www.runninghub.ai/post/2078038950391943170/?inviteCode=rh-v1603">https://www.runninghub.ai/post/2078038950391943170/?inviteCode=rh-v1603</a>Deeply fine-tuned on Krea2 Turbo, Edit_v1.0 perfectly retains ultra-fast inference and native aesthetics while specifically eliminating the "plastic" look in human portraits. It delivers a true photographic texture, enhancing the realism of skin, hair, light transitions, clothing, accessories, and background details. Furthermore, it integrates krea2-identity-edit capabilities, natively supporting Text-to-Image, Inpainting, and Image Reference for a fast, all-in-one creative experience.Note: This model's image editing is less precise and less specialized than native models such as Qwen-Image-Edit.<hr />2026-7-14中文介绍 / Chinese Description模型简介：SinoXvision_Krea2_Turbo_Uncen本模型基于 Krea2 Turbo 基础架构，针对超写实人像（Ultra-Realistic Portraiture）与高精细纹理（High-Definition Texture Enhancement）进行了深度微调。在保持原模型优秀的美学风格底蕴与全图生成稳定性的前提下，本版本重点攻克了生动皮肤质感、微小毛孔细节以及环境光影与人体贴合度等技术难点。核心技术改进（Technical Improvements）：现实增强（Enhanced Reality）： 深入修正了原生 Turbo 模型在复杂光影下人像容易出现的“塑料感”或“过度平滑”问题。通过定向微调，使眼球反光、皮肤纹理（如毛孔、微小瑕疵）、发丝边缘的扩散光表现更贴近真实物理世界。细节倍增（Enhanced Details）： 显著提升了中远景及复杂构图下的细节留存率。无论是衣物面料的经纬织纹、配饰的金属/材质光泽，还是背景中微小元素的清晰度，都得到了本质上的增强，且绝不产生画面过拟合的燥点。高速推理与稳定性： 完美继承了 Turbo 模型的极速响应特性。在不破坏原有构图框架与色彩美学的基础上，实现了画面精细度的平滑软着陆。推荐推理参数（Recommended Settings）：采样步数 (Steps): 8 - 12 （推荐 10 步以达到速度与细节的最佳平衡点）CFG Scale: 1.0 （请严格保持为 1，以确保 Turbo 架构的正常去噪与画面表现）采样器 (Sampler): Euler 或 Euler a / Euler_SDE提示词建议： 建议搭配高质量的摄影类词汇（如 cinematic lighting, photographic skin texture, 8k resolution）以彻底释放本模型的画质潜力。鸣谢（Acknowledgements）：特别感谢 salttaro 提供的 better_pussy LoRA。英文介绍 / English DescriptionModel Overview: SinoXvision_Krea2_Turbo_UncenThis model is a refined fine-tune based on the Krea2 Turbo architecture, meticulously optimized for Ultra-Realistic Portraiture and High-Definition Texture Enhancement. While preserving the inherent aesthetic DNA and structural stability of the base model, this version specifically addresses critical challenges in digital human rendering, such as lifelike skin tactile quality, microscopic pore detailing, and the integration of ambient lighting with human subjects.Key Features &amp; Enhancements:Enhanced Reality: It systematically resolves the "plasticy" or over-smoothed skin artifacts often found in native Turbo models under complex lighting conditions. The fine-tuning process yields vastly improved specular highlights in the eyes, realistic micro-textures on skin, and more natural light diffusion along hair strands.Enhanced Details: The model significantly improves detail retention in mid-to-long shots and intricate compositions. From the weave of textile fabrics and the material sheen of accessories to the clarity of subtle background elements, details are sharply elevated without introducing over-fitted noise or artifacts.Uncompromised Stability: It fully inherits the ultra-fast inference nature of the Turbo framework. The structural composition and color harmony of the original model remain uncompromised, ensuring a seamless upgrade to overall image fidelity.Recommended Inference Settings:Sampling Steps: 8 - 12 (10 steps is highly recommended for the optimal balance between speed and quality)CFG Scale: 1.0 (Strictly keep at 1.0 to ensure proper denoising behavior inherent to the Turbo architecture)Samplers: Euler or Euler_SDEPrompting Tips: Pair your prompts with high-quality photographic descriptors (e.g., cinematic lighting, photographic skin texture, 8k resolution) to fully unlock the potential of the enhanced details.Acknowledgements:Special thanks to salttaro for providing the better_pussy LoRA.

SinoX_Krea2 [ Aesthetics + Uncensored ]

该版本对数据序列进行了重新校准，在进一步强化人物真实感的同时，显著提升了画面美学质感与光影层次。相较于 SinoXvision‑krea2‑Uncen，该版本与原版 Turbo 的风格差异被进一步拉大。This version recalibrates the data sequence, further enhancing character realism while significantly elevating the visual aesthetics and lighting depth. Compared to SinoXvision‑krea2‑Uncen, the stylistic gap between this version and the original Turbo has been widened even further.

sinoxKrea2Aesthetics_visionV11Aesthetics.safetensors

sinoxKrea2Aesthetics_visionV11AesFP8.safetensors

Vision_v1.1(Aes FP8)

Vision_v1.1(Aes Int8)

Edit_v1.1_Int8_Convrot

Edit_v1.1

Edit_v1.0_Int8_Convrot

Edit_v1.0

v1.0(Uncen)_Int8

Vision_v1.0(Uncen)

该版本对数据序列进行了重新校准，在进一步强化人物真实感的同时，显著提升了画面美学质感与光影层次。相较于 SinoXvision‑krea2‑Uncen，该版本与原版 Turbo 的风格差异被进一步拉大。This version recalibrates the data sequence, further enhancing character realism while significantly elevating the visual aesthetics and lighting depth. Compared to SinoXvision‑krea2‑Uncen, the stylistic gap between this version and the original Turbo has been widened even further.

sinoxKrea2Aesthetics_visionV11AesInt8.safetensors

Files with SHA256: ab6e40a9ba26239bc448521ffd10e0d1c0f9a9544bf35f5cd48525d4042ab2ac