Downloads · 30 days
59
1% of all-time downloads
X-mix
les-chien/X-mix
X-mix is a text-to-image model from les-chien. Use it when you need an image from a text prompt. It is set up for diffusers. The card lists the license as creativeml-openrail-m.
Civitai: X-mix | Stable Diffusion Checkpoint | Civitai
All-time downloads
5.5K
Public
Repo size
36.7 GB
Likes
41
Public
Repo makeup
Click a slice to open those files.
.safetensors28.8 GB · 78%
At a glance
- Task
- Text-to-Image
- Library
- diffusers
- License
- creativeml-openrail-m
- Access
- Public
- Created
- Feb 21, 2023
- Updated
- Mar 18, 2023
- SHA
- 6c366cbf
Spaces
Model card
From the Hugging Face model README
X-mix
Civitai: X-mix | Stable Diffusion Checkpoint | Civitai
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.
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
Example
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
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
%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
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
%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
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
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)
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-Cogecha
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
masterpiece, best quality, ultra-detailed, illustration, portrait, 1girl
Negative prompt: EasyNegative, by bad-artist, bad_prompt_version2
Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 5, Seed: 1906918205, 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
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: 7bc4c05c90, Denoising strength: 0.55, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: Latent (bicubic antialiased), Eta: 0.67
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
Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 5, Seed: 3093571233, Size: 400x600, Model hash: 7bc4c05c90, Denoising strength: 0.55, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: Latent (bicubic antialiased), Eta: 0.67
1girl, on bed, wet, see-through shirt, thighhighs, cleavage, collarbone, full body,
Negative prompt: EasyNegative, photograph by bad-artist, bad_prompt_version2
Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 5, Seed: 1666118295, 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
Alice \(Alice in wonderland\), solo, Flowery meadow, cloudy sky, aqua eyes, white pantyhose, blonde hair,
Negative prompt: EasyNegative, sketch by bad-artist
Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 5, Seed: 807449917, Size: 512x512, Model hash: 7bc4c05c90, Denoising strength: 0.5, Clip skip: 2, ENSD: 31337, Hires upscale: 2, Hires upscaler: Latent (nearest-exact), Eta: 0.67
masterpiece, best quality, ultra-detailed, illustration, portrait, hakurei reimu, 1girl, throne room, dimly lit
Negative prompt: EasyNegative, by bad-artist, bad_prompt_version2
Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 5, Seed: 116927034, 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
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
Steps: 20, Sampler: DPM++ SDE Karras, CFG scale: 5, Seed: 1705759664, 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
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, 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
-
bad-artist: nick-x-hacker/bad-artist · Hugging Face
-
bad_prompt_version2: embed/bad_prompt · Hugging Face
-
Deep Negative V1.x: Deep Negative V1.x | Stable Diffusion TextualInversion | Civitai
You can consider whether to use them according to your preferences.
More
- 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.
- Using low resolution is not recommended for generating pictures.
- I did my best, but the hands are not perfect.
- The above settings may not necessarily be perfect.
- Due to my computer's performance, it's difficult for me to comprehensively test this model. I'm looking forward to your feedback.
- Task
- Text-to-Image
- Library
- diffusers
- License
- creativeml-openrail-m
- Created
- Feb 21, 2023
- Updated
- Mar 18, 2023