I just made a new Telegram Group, drop in for chats, questions, feedbacks or just to share your work. I will be sharing prompts and pics there as CIVITAI has been removing my work repeatedly. 要词加群。
All feedbacks are appreciated and will be considered for future releases!
Coffee? :P Ko-fi.com/catlover1937
2026-03-06
GGUF versions provided by @CH11442 , please show your thanks to him directly!
New Moody i2i Image2Image ZIT (Zimage Turbo) Simple Workflow - V1 available.
2026-03-05
Hello, HEADS UP, Moody V10 just dropped!
What’s new in V10?
In V10, I’ve baked in ZImage DPO, which enhances color vibrancy and lighting — areas where ZImage Turbo base model falls short, especially when running heavily loaded LoRAs, a lot of the colors are lost. If you’re interested in the technical approach behind it, I’ve included the reference link. I think this is a great addition that makes the outputs more eye-catching while still maintaining a natural look.
If you’d like even more vibrancy, you can also download the DPO LoRA and stack it on top of V10. Just make sure to experiment carefully and know when to stop — don’t go too crazy haha.
On top of the DPO integration, I’ve made a few minor image quality adjustments. I specifically tried to reduce the strange “liquid” effect on skin textures, which I personally always found annoying. It’s difficult to fully eliminate due to ZImage Turbo’s low-step nature, as fine details are harder to resolve, but I’ve reduced it as much as possible. As a result, you may notice the skin appears slightly more matte than before, giving it a more realistic, real-life texture.
V10 is compatible with the ZIT Simple workflow as well as the ZIB + ZIT workflow . A image to image ZIT workflow is in the work.
As always, stay safe, boys and girls. And your support is greatly appreciated ❤️
大家好，注意啦！V10 正式发布了。
V10 有什么新内容？
在 V10 中，我内置了ZImage DPO，它增强了色彩饱和度和光影表现——这正是 ZImage Turbo 原本的诟病，特别是在高负载 LoRA 情况下比较明显的色彩丢失的问题。如果你对技术实现方式感兴趣，我也附上了相关参考链接。我认为这是一次非常不错的融合，让画面在更吸睛的同时依然保持自然感。
如果你希望画面更加鲜艳，也可以额外下载 DPO LoRA 叠加在 V10 上使用。不过一定要自己多做实验，掌握好“适可而止”的程度，别叠得太猛哈哈。
除了 DPO 之外，我还对画质做了一些小调整。我尝试减少皮肤上那种奇怪的“液化感”纹理，这个问题我一直觉得很烦人。由于 ZImage Turbo 本身是低步数模型，细节收尾比较难彻底完成，所以很难完全消除，但我已经尽可能去优化。因此你可能会感觉皮肤质感比以前稍微更偏哑光一些，整体更接近真实生活中的皮肤效果。
和之前一样，V10 兼容 ZIT Simple 工作流以及 ZIB + ZIT workflow 工作流。
一如既往，大家注意安全。非常感谢你们一直以来的支持 ❤️
2026-02-22
Check out my new ZIB model with ZIB + ZIT workflow for ultimate details and variety, use with V9 for stunning effects.
全新 ZIB 模型发布！采用 ZIB + ZIT 工作流，带来极致细节与丰富变化。配合 V9 使用，视觉效果直接拉满。
GGUF versions (Q8, Q6, Q4) Provided by @prolma, please show your thanks directly to @prolma !
2026-02-21
Hello everyone,
V9 is here! Recently I’ve seen quite mixed feedback on V8. Some people feel it’s a big improvement, while others think V7 is more stable and visually pleasing. Because of that, I went back to the V7 foundation and rebuilt it into V9.
What I want to emphasize is that since V6, each version has had its own strengths and unique characteristics. A newer version isn’t automatically better than the previous one in every aspect. I encourage you to try them all and stick with the version that fits your taste and workflow best — there’s no need to always use the latest release.
If you prefer ultra-realistic rendering with highly detailed skin textures, you may also want to check out my experimental model, Moody Wild Mix.
Next, I’ll likely start working on releasing a non-distilled ZIB model to further refine and complete my "Moody ecosystem", especially the ZIB + ZIT workflow.
Thank you all for the continued support — and hopefully my GPU can keep fighting a little longer 😅
哈喽～
最近看到大家对 V8 的反馈有点两极分化，有人觉得提升很明显，也有人觉得 V7 更稳定、更耐看。所以这次我是在 V7 的基础上重新做了一个 V9。
其实从 V6 开始，每个版本都有自己的风格和侧重点，并不是新版本就一定全面碾压旧版本。大家可以都下载试试看，哪个顺手、哪个效果对胃口，就继续用哪个，不一定非要追最新版本。
另外，如果你偏好特别真实、皮肤细节很细腻那种风格，可以关注一下我的测试模型 Moody Wild Mix。
接下来我可能会开始研究并发布一个 ZIB 的未蒸馏模型，进一步完善我的 Moody 全家桶 ZIB + ZIT 工作流。
感谢大家一直以来的支持！也希望我的显卡还能再多战斗一阵子 😂
2026-02-13
Wondering where I’ve been this past month?
I’ve been retraining a bunch of my LoRAs on ZIB with my sh1tty GPU — and the compatibility turned out much better than I expected.
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 😏
I really hope you like it. If you enjoy my work, I’d truly appreciate your support — it would help me speed up future training and keep improving the model even further.
Edit 1: Forget to mention that, in V8 there was a different face model used than v7. Please let me know of your feedbacks.
想知道我这一个月在干嘛吗？
我一直在用我那张💩显卡，把一堆 LoRA 重新在 ZIB 上训练了一遍 —— 结果兼容性比之前好了很多。
姿势、脸部、细节——几乎所有你能想到的方面都有明显提升。我真的不知道为什么要一直折腾自己 😂 每次都说“这绝对是最后一个版本了”，但这个模型就是停不下来，越来越好 😏
希望你们会喜欢。如果你喜欢我的作品，也真的希望能得到你的支持，这会让我更快完成训练，继续把模型打磨得更好。
2026-01-28
check out my experimental workflow with ZIB X ZIT. A must try if you are getting bored of the predictablity of ZIT.
2026-01-26
Holy moly—v7 is finally here. v6 was already a strong foundation for what Moody X Zimage Turbo can do, so this release focuses mainly on fine-tuning: better faces, improved bodies, and more natural poses and lighting. That said… wow, making model eats so much time, especially with fine-tuning. I need to catch up on some real-life stuff 😅 Hopefully this will be the last model before the Zimage base is released.
2026-01-20
New simple workflow available in case you need one.
2026-01-16
Hi — V6 is finally here. I’ve been pushing hard to get this out as soon as possible, but it just didn’t feel ready. After a full week of endless testing, tweaking, and merging, I finally feel confident saying this is the best version so far — and a huge improvement over V5.
What’s new in V6:
Updated to use MysticXXX V5 in conjunction with V4, since V5 alone is quite destructive to the existing model
Removed several redundant LoRAs, or ones that only introduced disadvantages to the model
Rebased the model on the zImage Turbo base, making LoRA usage much easier and producing images that stay closer to the base model compared to V5
Huge improvement on genitalias
This version represents the most refined and stable state of the model to date. Hope you enjoy it!
Anyway, good luck generating! Feel free to share your results below—I’d love to see how others are using this model.
2026-01-09
After few sleepless nights and nonstop fine-tuning and testing, V5 IS HERE, extensive refinement and fine-detail adjustments were completed,
Found a working way to replace Mystic XXX V3 with V4.
Removes the unpleasant yellow color cast that affected outputs in V3.
Fixes pose anomalies when characters are farther from the camera.
Significantly improves XXX poses.
Further enhances compatibility with character LoRAs.
PS. Guys stop frigging' reporting my images, all my prompts start with 19 years old woman, and NOT friggin' minors. You know what, from now on you will only see me use "48 years old mature lady" instead :(
2026-01-07
Use V3 for consistency, V4 if you want to try something new. I am actively working on V5.
2026-01-06
HAPPY NEW YEAR! ZIT V4!
A lot of fine-tuning went into this version with the goal of improving compatibility with other LoRAs (especially character LoRAs), while preserving what V3 already does well. That said, please remember this is already an all-in-one checkpoint, so keep your LoRA weights very light. In my tests shown in the carousel, I used around 0.65 for character LoRAs and it worked quite well.
I experimented with MysticXXX V4, but it ended up breaking compatibility with other LoRAs. Once MysticXXX V4 is applied, adding additional LoRAs tends to cause distortions, so I decided to keep MysticXXX at V3 instead.
Baked in LSP to unlock more XXX poses 👏 big thanks to the creator, an absolute legend. However, One side effect is that it can drift quite a bit from the original faces. I’ve spent time trying to find a good balance, and I think the trade-off is worth it, but I’d love to hear what you think.
2025-12-25
Merry XMAS! ZIT V3 is here now
MysticXXX has updated their model to V3, so this release is now based on that version.
I’ve optimized the LoRA weights according to my preferences, with the hope that it’s more lightweight and compatible with additional LoRAs.
The FP8 version is also tested and up running OK.
PS: I’ve seen a few comments mentioning image breakage when stacking multiple LoRAs. I’ve noticed this in my own generations as well. That said, this checkpoint is an all-in-one, ready-to-use model, so you generally shouldn’t need to add extra LoRAs on top. Hopefully, with the V3 update and optimizations, images will be less distorted when LoRAs are used—but we’ll see.
Anyway, good luck generating! Feel free to share your results below—I’d love to see how others are using this model.
2025-12-24
Hi again after a bit more than a year,
I am annoyed of adding tones of lora every time when I generate so i merge this checkpoint.
I mixed RedCraft | 红潮造相 with turbo base as my base. Then merge DetailDeamonZ and [ZIT] Mystic XX PLUS 2 of the loras I trained, one gives better asian faces (to my own taste of course <3), nudes and nipples, one gives lighter skin (but not the pale and textureless one the turbo base has).
Why isn't there V1?
Because I never thought of sharing until someone asked, and I lost V1 :sadge:
Use at your own risk :)
PS. I am too lazy to add prompts, if you want them send me a DM.
SD 1.5
I was always looking for a model that can draw good nsfw photos with detailed private parts, so I made this for fun. Hope you like it.
This checkpoint generates very NSFW content. Put NSFW in the negative prompt if needed.
v4-beta: did some merging with my other model, and updated face model and smoothen the skin texture. Some people might like it while some don't, stick with v3 if you don't like the update.
2024-09-18: Re-uploaded the V3 pruned, seems to be working now. Sorry for the mess :(.
v3: Greatly improve model stability, better compactivities with NSFW Loras. Might look quite different from v2
Full/Pruned version available for v3
v2: Improved skin tone and quality of hands and faces as comment suggested.
Full/Pruned version available for v2
v1: initial release
Guidance: 7-15
Steps: 25 - 35
Sampler: DPM++ 2M
Hi-res fix 2x recommended. Adetailer on face was used.