Edit model card

You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this model content.


A stylized anime model. The goal of the mix is to affect the style of ink washed paintings (水墨画, lit. water rubbed art) popular in China, Japan, and Korea from the late 10th century until the modern day on anime illustrations.

Table of Contents

Model Descriptions


The Original. This is largely a proof of concept. It resembles most like GyozaMixV1.2 with a monochromatic filter on it and a prediliction to do hanfu or do asian style landscapes and scenary. Very controllable but the styllistic flavor is weak.



This is V1.0 but just remade with newer loras and a newer base model. I think it's better. It's more versatile at the very least.



Thanks to the aesthetic composition strength of the base model, GyozaMixV2, being strengthened it is possible to increase the styllistic flavor without it collapsing as easily. This model is very flavorful but is hard to control. Realistic tags will bring it back towards an anime style.



This model resembles anime more than V1.1 thanks to the addition of a watercolor and anime influences. However all its outputs are a bit generic looking. In addition, hand collapses even worse now. It is good at portrait shots and will perform subpar at most other things.



This model resembles the original SanXianWonton, but with big upgrades to color and landscape. It still slightly is monochromatic but is more of a generalized anime model now. In order to get it to resemble Shuimohua more closely it is advised that you use Moxin lora on it.



This model inherits the Silicon-29 license and is open access and available to all, with a CreativeML OpenRAIL-M license further specifying rights and usage. The CreativeML OpenRAIL License specifies:

  1. You can't use the model to deliberately produce nor share illegal or harmful outputs or content
  2. The authors claims no rights on the outputs you generate, you are free to use them and are accountable for their use which must not go against the provisions set in the license
  3. You may re-distribute the weights and use the model commercially and/or as a service. If you do, please be aware you have to include the same use restrictions as the ones in the license and share a copy of the CreativeML OpenRAIL-M to all your users (please read the license entirely and carefully) Please read the full license here :https://huggingface.co/spaces/CompVis/stable-diffusion-license

Terms of Use

  • Clearly indicate where modifications have been made.
    If you used it for merging, please state what steps you took to do so.


READ MORE: Disclaimer The user has complete control over whether or not to generate NSFW content, and the user's decision to enjoy either SFW or NSFW is entirely up to the user.The learning model does not contain any obscene visual content that can be viewed with a single click.The posting of the Learning Model is not intended to display obscene material in a public place. In publishing examples of the generation of copyrighted characters, I consider the following cases to be exceptional cases in which unauthorised use is permitted. "when the use is for private use or research purposes; when the work is used as material for merchandising (however, this does not apply when the main use of the work is to be merchandised); when the work is used in criticism, commentary or news reporting; when the work is used as a parody or derivative work to demonstrate originality." In these cases, use against the will of the copyright holder or use for unjustified gain should still be avoided, and if a complaint is lodged by the copyright holder, it is guaranteed that the publication will be stopped as soon as possible. I would also like to note that I am aware of the fact that many of the merged models use NAI, which is learned from Danbooru and other sites that could be interpreted as illegal, and whose model data itself is also a leak, and that this should be watched carefully. I believe that the best we can do is to expand the possibilities of GenerativeAI while protecting the works of illustrators and artists.


  • SanXianWonton.safetensors [c9d15393ae1cc461f1672f8edf1f33adc4ad91072a1ba59cd8baf8a42695daba]
  • 3FWonton+MSG.safetensors [fbdd29b923f0f1efe58ab6ec9b4152d15b068c4853430896ed069927aed67a83]
  • 3FWonton+anime.safetensors [ 2ecd09417d5a165f6c26accfcc58acd0fff313b6d63c116aa2fe3d781a3a053d]

Civitai: https://civitai.com/models/20330/three-delicacy-wonton-mix

Mixing Guide


  1. StoreGyozaV1.2 [7c8eef56]「9d44c42c2b630010a744f7c74a1185ff74ccd8b9a2f6359892996a92df3eec5b」
  2. GyozaMix2 [f8f043d1]「b04029407a6e59a2b6f5c355e8bac27e66cee2805d51835d72dec5f46e48c9c2」
  3. DPAnimelike 「d9454b0f315535607f8db94c0537ec465eff6d6ef262b0cee6fd1bc2a34e351a」


  1. KimHongdo 「f0d0d02c5eb4cf85441f4d104dc00b1510a0e35ded68e37804b5de863c0e3da1」
  2. Moxin 「3fd52c707c31b9af2207f697d9dd26b400683d011914471814bb33645f518f07」
  3. ShukezoumaV1.1 「f79768ec7b9e4f615458e0ea645424af183ffc0ebf020caab994eebe4dc84f7d」
  4. XiaorenshuV1.5 「dd9ead4035c17d3169fbb4a34a720b4909d54a5c367eaab5ae4b9e91eaebea3c」
  5. XiaorenshuV2 「50d66bdc578e56e368cb4756284313168b0b67af64db3305fdf30b7c67e37c1f」
  6. ShukezoumaColor 「cd115f0a3914a9547813566b7b9a0a8cb2bacb4aa8c9c4bedd400365ce9fcab5」
  7. WaterSmudge [ 51d5df24dbff32ff7ec203d1f237ac18d773722782ab2a7dde8882dc2510621b]
  8. DPanimelikeLora [ 23e6e35fb580a924eacd7b7bbd979cb431bdd99ab79543589bbb330541809277]
  9. Sili-landscape [52dd202159fb977e455bf57ba757a7afbcd118441f83192d3c9200a9f156d8f7]


SanXianWonton Recipe

Model Lora Weight Merge Name
StoreGyozaV1.2 KimHongdo 0.5 StoreGyozaV1.2_L
StoreGyozaV1.2_L Moxin 0.3 StoreGyozaV1.2_L2
StoreGyozaV1.2_L2 ShukezoumaV1.1 0.35 StoreGyozaV1.2_L3
StoreGyozaV1.2_L3 XiaorenshuV1.5 0.2 SanXianWontonV1

3FWonton Recipe

See above but replace XiaorenshuV1.5 with V2 and StoreGyozaV1.2 with GyozaMixV2.

3FWonton+MSG Recipe

Model Lora Weight Merge Name
GyozaMixV2 KimHongdo 0.5 GyozaMixV2_L
GyozaMixV2_L Moxin 0.5 GyozaMixV2_L2
GyozaMixV2_L2 ShukezoumaV1.1 0.7 GyozaMixV2_L3
GyozaMixV2_L3 ShukezoumaColor 0.6 GyozaMixV2_L4
GyozaMixV2_L3 XiaorenshuV2 0.2 3FWonton+MSG

3FWonton+anime Recipe

Model Lora Weight Merge Name
GyozaMixV2 KimHongdo 1.0 GyozaMixV2_L
GyozaMixV2_L ShukezoumaV1.1 0.7 GyozaMixV2_L2
GyozaMixV2_L2 XiaorenshuV2 0.5 GyozaMixV2_L3
GyozaMixV2_L3 Moxin 0.5 GyozaMixV2_L4
GyozaMixV2_L4 waterSmudge 0.8 GyozaMixV2_L5
GyozaMixV2_L5 DPanimelikeLora 1.0 3FWonton+anime

SanXianWonton2 Recipe

Model Lora Weight Merge Name
3FWonton+MSG sili-landscape 1.0 3FWonton+MSG+sili
Model: A Model: B Weight Base alpha Merge Name
DPAnimelike Silicon29-dark RCOSINE 0 DPAnimesili
Model: A Model: B Weight Base alpha Merge Name
DPAnimesili 3FWonton+MSG+sili 1,1,1,1,1,1,0,0,0,1,1,1,1,1,1,1,0,0,0,1,1,1,1,1,1 1 SanXianWonton2


SanXianWontonV1.0 Examples

Tall Aspect Ratio

SanXianWontonV1.1 Examples

Tall Aspect Ratio
Wide Aspect Ratio

SanXianWontonV1.2plusMSG Examples

Tall Aspect Ratio
Wide Aspect Ratio

SanXianWontonV1.3plusAnime Examples

Tall Aspect Ratio
Wide Aspect Ratio

SanXianWontonV2.0 Examples

Tall Aspect Ratio
Wide Aspect Ratio
Downloads last month
Inference API
Unable to determine this model's library. Check the docs .