Ani-il_v8/9.x
推荐配置|Recommended configuration：
Euler a, Schedule type: automatic
step: 36, CFG: 2.5, 832x1216/680x1208 Distilled CFG Scale:3.5
Hires.fix, Upscale by: 1.5|2 hires steps: 16, Denoising strength: 0.35, Hires CFG: 2.5, Hires Distilled CFG Scale:3.5
vae: 已自带
Ani-il_v7.1
推荐配置|Recommended configuration：
Euler a, Schedule type: automatic
step: 36, CFG: 5, 832x1216/680x1208 Distilled CFG Scale:3.5
Hires.fix, Upscale by: 1.5 hires steps: 16, Denoising strength: 0.35, Hires CFG: 3.5, Hires Distilled CFG Scale:3.5
vae: 已自带
推荐配合BlackStone HHP-SET_v5并使用下面提示词:
prompt:
__HHP-SET/Global/randomEroticArt__
OR
__HHP-SET/Global/randomExtreme__
negative_prompt:
__HHP-SET/Global/Negative__
Ani-il_v6.2
推荐配置|Recommended configuration：
Euler a, Schedule type: automatic
step: 36, CFG: 5, 832x1216/680x1208 Distilled CFG Scale:3.5
Hires.fix, Upscale by: 1.5 hires steps: 20, Denoising strength: 0.35, Hires CFG: 5, Hires Distilled CFG Scale:3.5
vae: 已自带
推荐配合BlackStone HHP-SET_v4并使用下面提示词:
prompt:
__HHP-SET/Global/randomEroticArt__
<lora:ILXL_Realism_Slider_v1 by klaabu @ civitai.com - thanks, cunt!:3.3> <lora:cknb02_stabilizer_v0.304a_fp16:0.35> <lora:DynamicPoseIL2att_alpha1.0_rank4_noxattn_900steps:1>
<lora:FLMFX:0.35> FLMFX, Film photo, Analog photography, 35mm film, Grainy texture, Film grain, Soft lighting, Vintage look, Washed colors, Light leaks, Natural shadows, <lora:Mirazhæl:0.35> miraz, scenery,
negative_prompt:
__HHP-SET/Global/Negative__
追加菜谱：
CatPony_Ani_il_v6.1=CatPony_Ani_il_v6+Ha9siro:1
CatPony_Ani_il_v6.2=CatPony_Ani_il_v6.1+HR_Giger_IL:1
Ani-il_v6
解开此版本封印的钥匙：
BlackStone HHP-SET_v1
BlackStone HHP-SET_v2
此版本比v5.1的背景更加清晰，背景光强度也有稍微下降，继承了v5.1优良的光影效果，能够轻松抽出令人惊艳的好图。比v5.1稍微不足的是，身姿倩影要稍微缺少那么一丝灵气（可能是我的主观感受）。
还有一个没有解决的问题是人像身边会莫名出现一个插有弓箭的箭筒。 ：（ 虽然还有解决这个问题，但这个版本却是最能让我眼前一亮的版本（相信不止一亮^_^）。希望大家能够玩的开心~~，最后，一定要试试BlackStone HHP-SET_v1，用过你就知道它的妙了，谁用谁知道~
此版本更新了两个重要的混合组成模型：
JANKU Trained + NoobAI + RouWei Illustrious XL v6.9
boleromix(illustrious) v9
DynamicPoseIL2att_alpha1.0_rank4_noxattn_900steps
模型混合菜谱如下：
CPA-il_v6_s0_sum50-50_v6=(boleromixIllustrious_v900 x 0.5 +waiIllustriousSDXL_v160 x 0.5) x 0.5 + JANKUTrainedNoobaiRouwei_v69 x 0.5
CPA-il_v6_s1_sm60_v6=CPA-il_v6_s0_sum50-50_v6.fp16 x 0.4 + oneObsession_v19 x 0.6
CPA-il_v6_s2_tr70=CPA-il_v6_s1_sm60_v6.fp16 + (BondageXL-2025-e61 - sd_xl_base_1.0) x 0.7_trainDifference
CPA-il_v6_s3_tr80=CPA-il_v6_s2_tr70.fp16 + (punishmentdiffusionxl_v07 - illustriousXL_v01) x 0.8_trainDifference
CPA-il_v6_s31_co60=CPA-il_v6_s3_tr200.fp16 x 0.4 + CPA-il_v6_s21_sm50.fp16 x 0.6_cosineA
CPA-il_v6_s4_i=CPA-il_v6_s31_co60+Expressive_H-000001:0.8,Dramatic Lighting Slider:5,People'sWorks_v9_Illusv1.0_128dim:0.35,MYRHA-Ψ00:0.35,Drävosk-:0.35,Vr4zkar:0.35,DynamicPoseIL2att_alpha1.0_rank4_noxattn_900steps:3
CPA-il_v6_s5_co50=CPA-il_v6_s4_i.fp16 x 0.5 + CatPony-Ani_il_v5.1.fp16 x 0.5_cosineA=CatPony_Ani_il_v6
推荐配置|Recommended configuration：
Euler a, Schedule type: automatic
step: 35, CFG: 3, 832x1216,（一定要够大，不然图片质量会很差）
Hires.fix, Upscale by: 1.5 hires steps: 16, Denoising strength: 0.5, Hires CFG: 3
vae: 已经自带
Ani-il_v5.1
CatPony_Ani_il_v5.1=CatPony_Ani_il_v5+Dramatic Lighting Slider:5,People'sWorks_v9_Illusv1.0_128dim:0.35,MYRHA-Ψ00:0.35,Drävosk-:0.35,Vr4zkar:0.35,2dRealPhoto_Background:0.35
Ani-il_v5.0
这个版本更新了所有混合模型到最新版本并重新混合。
感谢所有这些模型的制作人~~
下面是我的私房菜谱：
JANKU Trained + NoobAI + RouWei Illustrious XL v6
boleromix(illustrious) v8.15
oneObsession v19
WAI-illustrious-SDXL v16.0
BondageXL-2025-e61 (找不到访问地址了）
punishmentdiffusionxl_v07
模型混合工具：sd-webui-supermerger
混合工序：
CPA-il_v5_s0_sum50-50=(boleromixIllustrious_v815Cute x 0.5 +waiIllustriousSDXL_v160 x 0.5) x 0.5 + JANKUTrainedNoobaiRouwei_v60 x 0.5
CPA-il_v5_s1_sm60=CPD-il_v4_s0_sum50-50.fp16 x 0.4 + oneObsession_v19 x 0.6
CPA-il_v5_s2_tr200=CPD-il_v4_s1_sm60.fp16 + (BondageXL-2025-e61 - sd_xl_base_1.0) x 2_trainDifference
CPA-il_v5_s21_sm50=CPD-il_v4_s2_tr200.fp16 x 0.5 + CPD-il_v4_s1_sm60.fp16 x 0.5
CPA-il_v5_s3_tr200=CPD-il_v4_s21_sm50.fp16 + (punishmentdiffusionxl_v07 - illustriousXL_v01) x 2_trainDifference
CPA-il_v5_s31_co60=CPD-il_v4_s3_tr200.fp16 x 0.4 + CPD-il_v4_s21_sm50.fp16 x 0.6_cosineA
CatPony_Ani_il_v5=CPA-il_v5_s31_co60+Expressive_H-000001:0.8,add-detail-xl:1,zy_Detailed_Backgrounds_v1:1
最后一步融合了三个LORA：
zy_Detailed_Backgrounds_v1 (找不到访问地址了）
Ani-il_v3.0
去除了v2.1里黑暗元素，重新调整混合比例，色彩更鲜艳，画面更清晰。比较擅长2.5D，质感超赞~。~
The dark elements from v2.1 have been removed, and the color balance has been readjusted for more vivid colors and clearer images. It excels at 2.5D, with superb texture.
推荐配置|Recommended configuration：
Euler a, Schedule type: automatic
step: 60, CFG: 5, 684x1024,
Hires.fix, Upscale by: 1.5 hires steps: 20, Denoising strength: 0.35, Hires CFG: 3
vae: none
如何解开封印|How to break the seal？
去这里|Go here===> BlackStone V2
下载NSFW-DB.yaml文件至本地wildcards文件夹下，并刷新读取。|Download the NSFW-DB.yaml file to the local wildcards folder and refresh to read it.
如果不知道上面说的是什么，请看这里 | If you don't know what the above refers to, please see here：sd-dynamic-prompts
选择下面的提示词并生成图片~~！ Goodluck and have fun~
prompt:
例1|Example 1：
__NSFW-DB/prompt/prompt__,
white horse-like ears, white horse-like tail, red eyes, white hair,
({ballgag,drool,drooling|},metal collar, leash to the collar),dressed in skin-tight __NSFW-DB/Colors__ __NSFW-DB/Sexy Clothing__,
set in __NSFW-DB/Backgrounds/*__, filmed at __NSFW-DB/sexyShot/random__,
例2|Example 2：
__NSFW-DB/prompt/prompt__,
__NSFW-DB/prompt/autoRandom__
negative_prompt:
__NSFW-DB/prompt/negative_prompt__
===========================================================
The new CatPony-Dark series is here!
https://civarchive.com/models/1873150?modelVersionId=2120160
欢迎来到CatPony-Dark的新世界，虽然光与影总是相伴而生，但从融合模型这个角度上看，难度却增大了很多，特别是Dark需要融合一些独特的动画风格模型，导致模型的表现效果上总是差了一些（毕竟真实模型中很难描绘现实世界中不存在的事物），这也导致图片生成的非常不稳定。再三思考后，我还是决定把Dark单立一个新系列。再次欢迎来到与光对立的影之世界：CatPony-Dark
Welcome to the new world of CatPony-Dark. While light and shadow always coexist, the challenge becomes significantly greater when viewed from the perspective of fusion models, especially since Dark requires the integration of unique animation style models. This often results in suboptimal performance in model rendering (after all, it is difficult to depict non-existent elements from the real world in realistic models), leading to highly unstable image generation. After careful consideration, I have decided to establish Dark as a separate new series. Welcome once again to the shadow world that stands in opposition to light: CatPony-Dark.
封印解除的钥匙|The key to unlocking the seal：BlackStone V2
Ani-il_v2.0_Dark is here~~
黑暗之钥匙在推荐资源里----> 黑石
The Key to Darkness is in the recommended resources ----> Blackstone
https://civarchive.com/models/1833635?modelVersionId=2075016
祝大家玩得开心~~ Have fun!
Real_v3.1 = Real_v3.0 +中国元素
Real_v3.0 is here~~
更真实，更更更真实~~
Real_v2.1 is here~~
更好的身材，更出众的容貌，玩的开心~
眼睛有一些瑕疵 ：(
独乐乐不如众乐乐，如有可能，请多发图~
Better body, more outstanding looks, have fun~
Eyes have some blemishes :(
please post more pics if possible~
Real_v2.0 is here~~
More realism, more detail, incorporating more bondage elements
Real_v1.0 is here~
Pony + BDSM would be so much fun!
V2 has been released.
If you like the model, please post a few more pictures of your compliments~
Hello~ I hope this model can generate some content that couldn't be generated before, her birth has brought together the heart and efforts of many friends, and I hope that my efforts can bring some joy to all of you~
Special Thanks：
boleromix(Pony) - v1.41 VAE | Stable Diffusion Checkpoint | Civitai by unknown_bolero537
sdxl_99mix - v1.0 | Stable Diffusion Checkpoint | Civitai by 999999999636
I learned a lot of lessons and knowledge.
License: Fair AI Public License 1.0-SD
You should share the merge recipe if you release a model merged with pony_pencil-XL.
This license does not add any restrictions on your generated images.
For more details, please see the license section of ANIMAGINE XL 3.0 / Pony Diffusion V6 XL
Please see the "About this Version" section for information on each version.
Recommended Setting
Steps: 30
CFG scale: 7
Sampler: Euler a/DPM++ 2M (Karras/Automatic)
ADetailer face_yolov8n/s.pt use can fix eyes.(I don't really use it, but give it a try)
Positive Prompt
score_9, score_8_up, score_7_up,source_anime,BREAK,
...
<lora:detailed_notrigger:1><lora:Expressive_H:1><lora:naipf:0.8>
Negative Prompt
(worst,worse,average,source_furry,source_pony,simple background,blurry-background,bad anatomy,bad hands,missing fingers,:1.3)
Recipe
boleromix(Pony) - v1.41 VAE | Stable Diffusion Checkpoint | Civitai by unknown_bolero537
a.pony_pencil-XL - v2.0.0 | Stable Diffusion Checkpoint | Civitai by blue_pen5805
b.MomoiroPony - v1.4 | Stable Diffusion Checkpoint | Civitai by oritatami_neko
c.Mala Anime Mix NSFW PonyXL - v6.0 | Stable Diffusion Checkpoint | C... by Malary
i.WAI-ANI-NSFW-PONYXL - v6.0 | Stable Diffusion Checkpoint | Civitai by WAI0731
ii.Omega Pony XL anime style - v2.0 | Stable Diffusion Checkpoint | Ci... by kitarz
d.ChacolEbaraMixXL - v2.1 | Stable Diffusion Checkpoint | Civitai by chacols
e.EvaClausMix Pony XL - v2.1 | Stable Diffusion Checkpoint | Civitai by EvaClausAI
CPA_M1c=1.c.i + 1.c.ii
waiANINSFWPONYXL_v60 x 0.5 + omegaPonyXLAnime_v20 x 0.5_cosineA
CPA_M1A1 = 1.a + 1.b + CPA_M1c
(ponyPencilXL_v200 x 0.5 +momoiropony_v14 x 0.5) x 0.5 + CPA_M1c.fp16 x 0.5_cosineA
CPA_M1A2 = CPA_M1A + 1.d + 1.e
sdxl_99mix - v1.0 | Stable Diffusion Checkpoint | Civitai by 999999999636
a.ECLIPSE XL - v1.0 | Stable Diffusion Checkpoint | Civitai by Wasabiya
b.anima_pencil-XL - v5.0.0 | Stable Diffusion Checkpoint | Civitai by blue_pen5805
c.Animagine XL V3.1 - v3.1 | Stable Diffusion Checkpoint | Civitai by CagliostroLab
d.OsukiniMix XL - v1.0 | Stable Diffusion Checkpoint | Civitai by Vsukiyaki
e.LizMix - Version EX | Stable Diffusion Checkpoint | Civitai by hemlok
f.Copycat - v2.0 | Stable Diffusion Checkpoint | Civitai by calculater
g.ChacolOmegaMixXL - v1.1a | Stable Diffusion Checkpoint | Civitai by chacols
h.hibiscusmix - x522 | Stable Diffusion Checkpoint | Civitai by sakanaxbt
i.Pony Diffusion V6 XL - V6 (start with this one) | Stable Diffusion ... by PurpleSmartAI
WorkflowA：
CPA_M2A1=2.a+2.b+2.c
(eclipseXL_v10 x 0.5 +animaPencilXL_v500 x 0.5) x 0.5 + animagineXLV31_v31 x 0.5_cosineA
CPA_M2A2=CPA_M2A1+2.d+2.e
(CPA_M2A1.fp16 x 0.5 +osukinimixXL_v10 x 0.5) x 0.5 + lizmix_versionEX x 0.5_cosineA
CPA_M2A3=CPA_M2A2+2.f+2.g
(CPA_M2A2.fp16 x 0.5 +copycat_v20 x 0.5) x 0.5 + chacolomegamixxl_v11a x 0.5_cosineA
CPA_M2A4=CPA_M2A3+2.h
CPA_M2A3.fp16 x 0.5 + hibiscusmix_x522 x 0.5_cosineA
WorkflowB：
CPA_M2B1=2.a+(2.b-2.c)
eclipseXL_v10 + (animaPencilXL_v500 - animagineXLV31_v31) x 1_trainDifference
CPA_M2B2=CPA_M2B1+(2.d-2.c)
CPA_M2B1.fp16 + (osukinimixXL_v10 - animagineXLV31_v31) x 1_trainDifference
CPA_M2B3=CPA_M2B2+(2.e-2.c)
CPA_M2B2.fp16 + (lizmix_versionEX - animagineXLV31_v31) x 1_trainDifference
CPA_M2B4=CPA_M2B3+(2.f-2.i)
CPA_M2B3.fp16 + (copycat_v20 - ponyDiffusionV6XL_v6StartWithThisOne) x 1_trainDifference
CPA_M2B5=CPA_M2B4+(2.g-2.i)
CPA_M2B4.fp16 + (chacolomegamixxl_v11a - ponyDiffusionV6XL_v6StartWithThisOne) x 1_trainDifference
CPA_M2B6=CPA_M2B5+(2.h-2.i)
CPA_M2B5.fp16 + (hibiscusmix_x522 - ponyDiffusionV6XL_v6StartWithThisOne) x 1_trainDifference
CPA_M2B7=CPA_M2B6+2.a
CPA_M2B6.fp16 x 0.5 + eclipseXL_v10 x 0.5_cosineA
CPA_M2B7A=CPA_M2B6+CPA_M2A4
CPA_M2B6.fp16 x 0.5 + CPA_M2A4 x 0.5_cosineA
CatPony_Ani_v10=CPA_M1A2 + CPA_M2B7A
CPA_M1A2.fp16 x 0.5 + CPA_M2B7A_M2B6+M2A4.fp16 x 0.5_cosineA
extremely detailed (no trigger) - sliders.ntcai.xyz - v2.0 | Stable... by ntc
ExpressiveH (Hentai LoRa Style) エロアニメ - ExpressiveH | Stable Diffus... by jspp3733410
nai style for pony - final | Stable Diffusion LyCORIS | Civitai by ComprehensivePrize65844
Baked VAE
SharpSpectrumVAEXL - v1 | Stable Diffusion VAE | Civitai by Pitpe11
Description
FAQ
Comments (10)
CatPony Real-IL v1.0 doesn't load at all and only gives the following error (根本无法加载，只出现以下错误):
"in fail_if_models_did_not_load raise Exception(f"Could not load the {model_type} model! Reason: " + e) Exception: Could not load the stable-diffusion model! Reason: 'Embedding' object has no attribute 'wrapped'
Didn't have this problem with previous versions (以前的版本没有这个问题).
我还没遇到过类似的情况，可以考虑重新下载一次试试看
I'll try re-downloading it again, but I think it's complete (我会尝试重新下载，但我认为它已经完整了). My file size is (我的文件大小为) 6.64 GB = 7,140,418,106 bytes. Never had a problem loading any other models (加载任何其他模型从未遇到问题).
AFD_0 如果重新下载完还有同样的问题，可以尝试重启下电脑。我以前遇到过几次，运行失败的情况，重启之后就正常了。
ggyydream Absolutely, but probably not the problem. Since then I've loaded several other models and they all work fine. I'll try to re-download this week. Thanks! (绝对，但可能不是问题。从那时起，我已经加载了其他几个型号，它们都可以正常工作。我将尝试本周重新下载。谢谢!)
try real-il_v2.0
@ggyydream I will, thank you! (我会的，谢谢).
@ggyydream No luck. Same error message "Error: Could not load the stable-diffusion model! Reason: 'Embedding' object has no attribute 'wrapped'." Never saw this error message before out of hundreds of other checkpoints, other than on CatPony Real-IL v1.0 and Real-IL v2.0 (运气不好。同样的错误信息：“错误：无法加载稳定扩散模型！原因：‘嵌入’对象没有‘wrapped’属性.” 除了 CatPony Real-IL v1.0 和 Real-IL v2.0 之外，在数百个检查点中，从未见过此错误信息)
@AFD_0 我让Grok看了下这个错误报告，他给出的分析是这样的，你看看是否有帮助。
The error you're encountering, "Could not load the stable-diffusion model! Reason: 'Embedding' object has no attribute 'wrapped'", likely stems from a mismatch in how an Embedding object is being handled during model loading in a Stable Diffusion setup (possibly using Hugging Face Diffusers or a related library like LangChain for embeddings). This could be due to version incompatibilities, incorrect input types, or issues with model wrapping (e.g., in distributed training or custom embeddings like textual inversion).
Possible Causes
Object Mismatch: The code expects the Embedding object to have a 'wrapped' attribute (possibly from a wrapped module in frameworks like PEFT or FSDP), but it's treating it as a standard torch.nn.Embedding or similar, which doesn't have that attribute.
Version Issues: Outdated versions of diffusers, transformers, or torch can cause attribute errors when loading models or embeddings.
Input Type Error: If using embedding functions (e.g., HuggingFaceEmbeddings.embed_documents), the input might be Document objects instead of strings, leading to attribute errors during processing.
Virtual Environment Corruption: The virtual environment (venv) might be corrupted or built with incompatible Python versions, causing object attribute issues.
Solutions
Update Libraries: Ensure you're using the latest versions, as attribute errors are often fixed in updates.
Run:
text
pip install --upgrade diffusers transformers torch
This has resolved similar attribute errors in Diffusers pipelines.
Check Input Types for Embeddings: If this occurs during embedding generation (e.g., textual inversion or HuggingFaceEmbeddings), ensure inputs are strings, not objects like Document. For example, in LangChain:
Use:
text
doc_func = lambda x: x.page_content
docs = list(map(doc_func, your_documents))
embeddings = embedding_model.embed_documents(docs)
This extracts string content and avoids attribute errors on objects.
Recreate Virtual Environment: Delete and recreate the venv to fix corruption or version mismatches.
Steps:
Delete the venv folder in your Stable Diffusion directory.
Re-run the setup script (e.g., webui-user.bat for A1111) to recreate it.
This fixed similar initialization errors like 'str' no 'isascii'.
Launch Flags for Stable Diffusion: If using AUTOMATIC1111's WebUI or similar:
Add flags like --no-half or --precision full to webui-user.bat or command line to avoid half-precision issues, which can cause model loading errors. Example: Set COMMANDLINE_ARGS= --no-half in the bat file.
Check Model/Embedding Files:
Ensure embeddings are correctly placed (e.g., in embeddings/ folder for A1111) and have .pt or .bin extension.
If using custom embeddings, test without them to isolate the issue.
Redownload the Stable Diffusion model if corrupted.
If these don't resolve it, provide more context (e.g., code snippet, library versions, or full traceback) for further diagnosis. If it's in a specific framework like Diffusers, check the GitHub issues for recent bugs
@ggyydream >>issues with model wrapping (e.g., in distributed training or custom embeddings like textual inversion)
Like I mentioned, I've only ever experienced this error with 2 of your models out of hundreds of other SDXL-compatible checkpoints I've been testing that load just fine. The only other ones that are unsupported by my software setup is FP8 and Flux. FP16, BF16, FP32 versions of SDXL, Illu and Pony all work. No big deal, might just be an incompatibility somewhere. Thanks for helping to troubleshoot, appreciate that.
>>模型包装的问题（例如，在分布式培训或自定义嵌入中，例如文本反演）
就像我提到的那样，我已经测试了数百个与SDXL兼容的检查点，并且全部正确加载。只有Catpony Real-IL V1和V2有这个问题。我的软件不支持的唯一其他事情是FP8和Flux。 FP32，FP16，BF16，SDXL，Illu和Pony都得到了支持。没什么大不了的。也许它只是不兼容的地方。感谢您尝试帮助解决问题