Update README.md

README.md

@@ -5,21 +5,139 @@ tags:
 - stable-diffusion
 pipeline_tag: text-to-image
 ---
-X-mix is a merging model used to generate anime images. I have forgotten the recipe for X-mix, as too many models were used for merging. My English is not very good, so there may be some parts of this article that are unclear. This model supports nsfw.
+# X-mix
+**Civitai**: [X-mix | Stable Diffusion Checkpoint | Civitai](https://civitai.com/models/13069/x-mix)
 
-## How to use
+X-mix is a merging model used to generate anime images. My English is not very good, so there may be some parts of this article that are unclear. 
 
-- Recommended Negative Prompt: **EasyNegative, by bad-artist, bad_prompt_version2**. Related links: [embed/EasyNegative · Hugging Face](https://huggingface.co/embed/EasyNegative) [nick-x-hacker/bad-artist · Hugging Face](https://huggingface.co/nick-x-hacker/bad-artist) [embed/bad_prompt · Hugging Face](https://huggingface.co/embed/bad_prompt)
-- Recommended Sampler: DPM++ SDE Karras
-- Recommended Steps: 20
-- Recommended CFG Scale: 5
-- Recommended Hires upscaler: Latent (bicubic antialiased), Denoising strength: 0.5~0.6
-- Recommended vae: orangemix.vae.pt Related link: [VAEs/orangemix.vae.pt · WarriorMama777/OrangeMixs at main (huggingface.co)](https://huggingface.co/WarriorMama777/OrangeMixs/blob/main/VAEs/orangemix.vae.pt) 
+## V2.0
+
+V2.0 is a merged model based on V1.0. This model supports nsfw. 
+
+### Difference from V1.0
+
+- The performance of V2.0 is not better than that of V1.0, but the generated images now exhibit a different artistic style.
+- V2.0 offers better support for nsfw than V1.0, but the drawback is that even when you do not intend to generate an nsfw image, there is still a possibility of generating one. If you are more interested in the sfw model, I will provide a detailed explanation in the recipe section.
+- In my opinion, V2.0 is not as user-friendly as V1.0, and it appears to be more challenging to generate an excellent image.
+
+### Recommended Settings
+
+- Sampler: DPM++ SDE Karras (sfw), DDIM (nsfw)
+- Steps: 20 (DDIM may require more steps)
+- CFG Scale: 5
+- Hires upscale: Latent (bicubic antialiased), Latent (nearest-exact), Denoising strength: 0.4~0.7
+- vae: NAI.vae
 - Clip skip: 2
 - ENSD: 31337
 - Eta: 0.67
 
-## Examples
+### Example
+
+![](https://raw.githubusercontent.com/les-chien/MyGallery/markdown/pictures/03229-4291846267-masterpiece%2C%20best%20quality%2C%20ultra-detailed%2C%20illustration%2C%20portrait%2C%201girl.png)
+
+```
+masterpiece, best quality, ultra-detailed, illustration, portrait, 1girl
+Negative prompt: EasyNegative, photograph by bad-artist, bad_prompt_version2, DeepNegative-V1-75T
+Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 5, Seed: 4291846267, Size: 512x512, Model hash: 7bc4c05c90, Denoising strength: 0.55, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: Latent (nearest-exact), Eta: 0.67
+```
+
+![](https://raw.githubusercontent.com/les-chien/MyGallery/markdown/pictures/03003-2118045521-Indoor%2C%20bright%2C%201Girl%2C%20gray%20hair%2C%20amber%20eyes%2C%20smile%2C%20black%20dress%2C%20barefoot%2C%20sitting%20posture%2C.png)
+
+```
+Indoor, bright, 1Girl, gray hair, amber eyes, smile, black dress, barefoot, sitting posture,
+Negative prompt: EasyNegative, by bad-artist, bad_prompt_version2
+Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 5, Seed: 2118045521, Size: 600x400, Model hash: 7961a4960e, Denoising strength: 0.55, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: Latent (bicubic antialiased), Eta: 0.67
+```
+
+![](https://raw.githubusercontent.com/les-chien/MyGallery/markdown/pictures/03007-3093571233-landscape%2C%20in%20spring%2C%20cherry%20blossoms%2C%20cloudy%20sky%2C%201girl%2C%20solo%2C%20long%20blue%20hair%2C%20smirk%2C%20pink%20eyes%2C%20(school%20uniform_1.05)%2C%20white%20t.png)
+
+```
+landscape, in spring, cherry blossoms, cloudy sky, 1girl, solo, long blue hair, smirk, pink eyes, (school uniform:1.05), white thighhighs,
+Negative prompt: EasyNegative, by bad-artist, bad_prompt_version2, DeepNegative-V1-75T
+Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 5, Seed: 3093571233, Size: 400x600, Model hash: 7961a4960e, Denoising strength: 0.55, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: Latent (bicubic antialiased), Eta: 0.67
+```
+
+![](https://raw.githubusercontent.com/les-chien/MyGallery/markdown/pictures/03009-986400693-1girl%2C%20on%20bed%2C%20wet%2C%20see-through%20shirt%2C%20thighhighs%2C%20cleavage%2C%20collarbone%2C%20full%20body%2C.png)
+
+```
+1girl, on bed, wet, see-through shirt, thighhighs, cleavage, collarbone, full body,
+Negative prompt: EasyNegative, photograph by bad-artist, bad_prompt_version2, DeepNegative-V1-75T
+Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 5, Seed: 986400693, Size: 512x512, Model hash: 7961a4960e, Denoising strength: 0.55, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: Latent (bicubic antialiased), Eta: 0.67
+```
+
+![](https://raw.githubusercontent.com/les-chien/MyGallery/markdown/pictures/03011-273840053-Alice%20_(Alice%20in%20wonderland_)%2C%20solo%2C%20Flowery%20meadow%2C%20cloudy%20sky%2C%20aqua%20eyes%2C%20white%20pantyhose%2C%20blonde%20hair%2C.png)
+
+```
+Alice \(Alice in wonderland\), solo, Flowery meadow, cloudy sky, aqua eyes, white pantyhose, blonde hair,
+Negative prompt: EasyNegative, photograph by bad-artist, bad_prompt_version2, DeepNegative-V1-75T
+Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 5, Seed: 273840053, Size: 512x512, Model hash: 7961a4960e, Denoising strength: 0.55, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: Latent (bicubic antialiased), Eta: 0.67
+```
+
+![](https://raw.githubusercontent.com/les-chien/MyGallery/markdown/pictures/03005-2212365348-masterpiece%2C%20best%20quality%2C%20ultra-detailed%2C%20illustration%2C%20portrait%2C%20hakurei%20reimu%2C%201girl%2C%20throne%20room%2C%20dimly%20lit.png)
+
+```
+masterpiece, best quality, ultra-detailed, illustration, portrait, hakurei reimu, 1girl, throne room, dimly lit
+Negative prompt: EasyNegative, by bad-artist, bad_prompt_version2, DeepNegative-V1-75T
+Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 5, Seed: 2212365348, Size: 512x512, Model hash: 7961a4960e, Denoising strength: 0.55, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: Latent (nearest-exact), Eta: 0.67
+```
+
+![](https://raw.githubusercontent.com/les-chien/MyGallery/markdown/pictures/03023-293615512-masterpiece%2C%20best%20quality%2C%20ultra-detailed%2C%20illustration%2C%201girl%2C%20witch%20hat%2C%20purple%20eyes%2C%20blonde%20hair%2C%20wielding%20a%20purple%20staff%20bla.png)
+
+```
+masterpiece, best quality, ultra-detailed, illustration, 1girl, witch hat, purple eyes, blonde hair, wielding a purple staff blasting purple energy, purple beam, purple effects, dragons, chaos
+Negative prompt: EasyNegative, photograph by bad-artist, bad_prompt_version2, DeepNegative-V1-75T
+Steps: 20, Sampler: DDIM, CFG scale: 5, Seed: 293615512, Size: 512x512, Model hash: 7961a4960e, Denoising strength: 0.55, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: Latent (nearest-exact)
+```
+
+![](https://raw.githubusercontent.com/les-chien/MyGallery/markdown/pictures/03242-3442031040-1girl%2C%20solo%2C%20black%20skirt%2C%20blue%20eyes%2C%20electric%20guitar%2C%20guitar%2C%20headphones%2C%20holding%2C%20holding%20plectrum%2C%20instrument%2C%20long%20hair%2C%20%2C%20mu.png)
+
+
+```
+1girl, solo, black skirt, blue eyes, electric guitar, guitar, headphones, holding, holding plectrum, instrument, long hair, , music, one side up, pink hair, playing guitar, pleated skirt, black shirt, indoors
+Negative prompt: EasyNegative, photograph by bad-artist, bad_prompt_version2, DeepNegative-V1-75T
+Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 5, Seed: 3442031040, Size: 512x512, Model hash: 7961a4960e, Denoising strength: 0.6, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: Latent (nearest-exact), Eta: 0.67
+```
+
+### Recipe
+
+**Step 1: **animefull-latest (model) + pastelmix-lora (lora) + ligneClaireStyleCogecha (lora) = pastel-Cpgecha
+
+You can try replacing animefull-latest with Anything-V3.0 or your preferred model. However, I cannot confirm if this will yield better results and it requires you to experiment with it on your own. 
+
+**Step 2: **MBW: Chilloutmix + X-mix-V1.0
+
+| Model A     | Model B    | base_alpha | Weight                                            | Merge Name      |
+| ----------- | ---------- | ---------- | ------------------------------------------------- | --------------- |
+| Chilloutmix | X-mix-V1.0 | 1          | 1,1,1,1,1,1,1,1,0,0,0,0,1,0,0,0,0,1,1,1,1,1,1,1,1 | X-mix-V2.0-base |
+
+This is the step of the sfw version. The steps for the nsfw version are as follows: I merged several LoRAs into Chilloutmix to obtain Chilloutmix-nsfw. Then I merged Chilloutmix-nsfw and X-mix-V1.0 to get X-mix-V2.0-nsfwBase1. Finally, I merged several LoRAs into X-mix-V2.0-nsfwBase1 to get X-mix-V2.0-nsfwBase2.
+
+LoRAs related to real people should be merged into Chilloutmix or other photo-realistic models that you like, while LoRAs related to anime should be merged into X-mix-V2.0-base. Which LoRAs to use depends on your preference.
+
+**Step 3: **MBW: pastel-Cogecha + X-mix-V2.0-base
+
+| Model A        | Model B         | base_alpha | Weight                                                  | Merge Name     |
+| -------------- | --------------- | ---------- | ------------------------------------------------------- | -------------- |
+| pastel-Cogecha | X-mix-V2.0-base | 0          | 1,1,1,1,1,0.3,0,0,0,1,0.1,1,1,1,1,1,0,1,0,1,1,0.2,1,1,1 | X-mix-V2.0-sfw |
+
+In fact, I never tried to obtain the sfw version because I didn't plan on using it from the beginning. So this process is for reference only, and I am not sure about the actual effect of the sfw model.
+
+## V1.0
+
+I have forgotten the recipe for X-mix-V1.0, as too many models were used for merging. This model supports nsfw, but the effect may not be very good.
+
+### Recommended Settings
+
+- Sampler: DPM++ SDE Karras
+- Steps: 20
+- CFG Scale: 5
+- Hires upscaler: Latent (bicubic antialiased), Denoising strength: 0.5~0.6
+- vae: NAI.vae
+- Clip skip: 2
+- ENSD: 31337
+- Eta: 0.67
+
+### Examples
 
 ![](https://raw.githubusercontent.com/les-chien/MyGallery/markdown/pictures/1girl.png)
 
@@ -85,6 +203,20 @@ Negative prompt: EasyNegative, by bad-artist, bad_prompt_version2
 Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 5, Seed: 2548407675, Size: 512x512, Model hash: 7bc4c05c90, Denoising strength: 0.55, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: Latent (bicubic antialiased), Eta: 0.67
 ```
 
+## Embedding
+
+If you need the embedding used in examples, click them:
+
+- **EasyNegative: **[embed/EasyNegative · Hugging Face](https://huggingface.co/embed/EasyNegative)
+
+- **bad-artist: **[nick-x-hacker/bad-artist · Hugging Face](https://huggingface.co/nick-x-hacker/bad-artist)
+
+- **bad_prompt_version2: **[embed/bad_prompt · Hugging Face](https://huggingface.co/embed/bad_prompt)
+
+- **Deep Negative V1.x: **[Deep Negative V1.x | Stable Diffusion TextualInversion | Civitai](https://civitai.com/models/4629/deep-negative-v1x)
+
+You can consider whether to use them according to your preferences.
+
 ## More
 
 1. Since my prompts are usually brief, I'm not sure if this model will be able to meet all of your requirements if you need to use a large number of prompts.