[Checkpoint] Ether Blu Mix

模型介绍：
🔹☁️ Hello, this is Ether Blu Mix ☁️🔹
Ether Blu Mix is a stylized anime model specialized in creating character illustrations with vibrant colors, expressive backgrounds and a painterly feel. Based on Silicon29 and Counterfeit 3, I consider this my interpretation of a pastel-mix style of model.

✨ Please Share Your Cool Creations Below! ✨

* Version 7 Update:
Ether Blu 7, while being largely a weights focused change, looks to further enhance the lighting focused, slightly niji feel of EBM 6 as well as its painterly origins.

Might be a little overcooked. Some prompts and generations, particularly those with heavier quality related keywords and negative prompts, are more likely to experience this. Please use the recommended negative prompts to help alleviate it.

VAE is baked, same one I've always preferred.

Thanks everyone, have fun. I'll be looking forward to your feedback!

* Huggingface Archive

☄️ Generation Recommendations
Use whatever sampler, steps, CFG, clip-skip you prefer.
Sample images were generated with
▣ Sampler: DPM++ 2M Karras
▣ Steps: 20 - 28
▣ CFG: 7 - 11
▣ Clip Skip: 1
▣ ENSD: 31337

Increasing CFG can lead to more vibrant results.

Negative Prompt Example:
"NSFW, (worst quality, low quality:1.3)"

I recommend this simple negative prompt if you wish to achieve a similar style to the preview images. You can further alter positive and negative prompts based on the type of style you'd like to achieve.

Style focused negative prompt examples:
* Photorealism feel:
▣ "sketch, drawing, anime, cartoon, illustration, NSFW, (worst quality, low quality:1.3)"
* Flat Anime/Cartoon:
▣ "realistic, photorealistic, photograph, 3D, NSFW, (worst quality, low quality:1.3)"

Recommended VAE:
The VAE is baked:
Waifu Diffusion VAE kl-f8-anime2

Strongly recommended as the model was tuned with this VAE's color and saturation in mind.
Sample images were further upscaled using AUTO1111 Hi-res fix. Please utilize it if you wish to achieve similar results.
▣ Latent (nearest-exact)
▣ Upscale by: 2
▣ Denoising strength: 0.55

Are your results changing too much after Hi-res fix? Please try 4x-foolhardy-Remarci.
(Also recommended for simple style illustrations, anime, cartoon, etc.)
▣ 4x-foolhardy-Remarci
▣ Upscale by: 2
▣ Denoising strength: 0.45 - 0.55

🖊️ Prompting Advice
The model was mixed and tested using these prefixes and will give you the most consistent style results. Of course you're free to prompt however you'd like as booru-style models tend to be quite flexible in prompting.

If you look at the prompts included with the sample images they demonstrate that even with a very loose and vague style of prompting you can achieve fun, high quality images.

Experiment and give the AI some freedom to interpret your ideas!

General guidelines:
▣ Prompt in the style of anime/booru-style models, "masterpiece, best quality, ..... "
▣ ▣ More anime style images
▣ Prompt in the style of photography/non-anime models, "professional photograph of ..... "
▣ ▣ More realistic images, probably higher style variety

📘 Model Merging Info - v7
I have no training or real knowledge of machine learning. I merged the model using the information provided on WarriorMama777's AOM model page as well as the info from these Rentrys. I very arbitrarily mixed the UBM weights of each model, modifying the values with trial and error until I was visually satisfied with it.

I utilized the SuperMerger extension for AUTO1111 by hako-mikan.

Please refer to the 'About This Version' for more detailed information.

💙 Credits
Thank you to WarriorMama777 for providing AOM and other various mixes as well as detailing your workflow and inspiring me to try mixing my own models.
Thank you to Luna for sharing your process on creating Kenshi and Aurora. It deepened my understanding of model making as well as inspired me to mix LORA's.
Thank you to hako-mikan for creating the SuperMerger extension allowing a quicker workflow for merging.
Thank you to gsdf, rqdwdw for Counterfeit.
Thank you to Xerxemi and Xynon for SD-Silicon.
Thank you to YozoRaAru for Night Sky YOZORA and Color Box.
Thank you to DynamicWang for AWPainting.
Thank you to Jemnite for Lunar Radiance.
Thank you to CTD for Dpep4, Xlimo and lightingv2.
Thank you to dolphinz for Sttabi.
Thank you to SatyamSSJ10 for Jordan 3.
Thank you to ashen-sensored for breakA lighting isolation.
Thank you to Terry9993 for Niji Default style.
Thank you to those at the LMFResearch Society for their archive of SD LORA's and checkpoints.
Thank you to AUTO1111 for creating the web UI.
Thank you to StabilityAI for starting everything with Stable Diffusion.
Thank you to CivitAI and Huggingface for providing places where open source AI can flourish.
Thank you to the entire SD community for continuing to openly share and create.

~~~~~~~~~~~~~~~~~~~~~~~o(*￣︶￣*)o~~~~~~~~~~~~~~~~~~~~~~~

示例提示词和参数：
Prompt: masterpiece, best quality, movie still, 1girl, floating in the sky, cloud girl, cloud, (close-up:1.1), bright, smug, smirk, happy, fun, soft lighting, volumetric lighting, Bauhaus, shapes, lines, sparkling, glittering, swirling, abstract, expressive

Negative prompt: NSFW, (worst quality, low quality:1.3)

Steps: 22, CFG scale: 9, Sampler: DPM++ 2M Karras, Seed: 2041342887, VAE: kl-f8-anime2.vae.pt, Size: 512x896, Model: EBM-7-LORAmerge-final-fp16, Version: v1.7.0, VAE hash: 551eac7037, Model hash: 686eec2aef, Hires upscale: 2, Hires upscaler: Latent (nearest-exact), Denoising strength: 0.55