Sign In

Ether Real Mix

Checkpoint Merge
Sep 6, 2023
Base Model
SD 1.5
Usage Tips
Clip Skip: 1

πŸ”ΈπŸŒ€οΈ Hello, this is Ether Real Mix πŸŒ€οΈπŸ”Έ

Ether Real Mix is a stylized realism model focused on flexibility. It's capable of producing a variety of subjects in multitudes of styles. It still has some shortcomings inherit to anime models such as light biases in generating females and humans.

This model is intended to act as a 'blank canvas'. Add your favorite LORA's and embeddings to customize it to your liking!

✨ Please Share Your Cool Creations Below! ✨

* Ether Real Mix - Version 3.1

Ether Real 3.1 brings significant improvements to style variety, as well as less bias and a refined base style.

3.1 serves as a turning point for Ether Real as I've finally been able to achieve better style flexibility. Previous versions, while quite flexible in terms of subject matter, were biased towards the standard 2.5D style. You typically had to heavily weight and prompt tags to push Ether Real into different styles. By significantly altering a lot of weights during MBW merging, focusing on adding more composition specific weights within more diverse models, Ether Real has much better style variety while retaining its base 2.5D feel and general purpose capabilities.

I substituted Deliberate with RunDiffusion FX Photorealistic and updated Juggernaut to Aftermath. While Deliberate is a great, diverse model I wanted to push Ether Real's photorealistic capabilities while complimenting Juggernaut Aftermath's style.

* TLDR: Better style variety, less biased, Papa John's.

Ether Real Mix 3.1 should be significantly easier to produce different styles, from a photorealistic feel, to simple anime/cartoon illustrations.

Deliberate 2 --> RunDiffusion FX Photorealistic

Juggernaut Final --> Juggernaut Aftermath

* Huggingface Archive

β˜„οΈ Generation Recommendations

* All preview images were generated with no LORA's, embeddings, extensions or img to img/inpainting. Just pure text to image with hi-res fix upscaling.

Use whatever sampler, steps, CFG you prefer.

Most sample images were generated with:

  • Sampler: DPM++ 2M Karras

  • Steps: 20 - 28

  • CFG: 7 - 11

  • Clip Skip: 1

  • ENSD: 31337

Higher 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)"

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

Waifu Diffusion VAE kl-f8-anime2

Strongly recommended as the model was tuned with this specific VAE's colors and saturation in mind.

πŸ–ŠοΈ Prompting Advice

Prompt however you like. The model was merged with both Danbooru-style tag prompting as well as natural language in mind. You're free to mix and match prompting styles, the model should understand it.

  • Generating unprompted girls/humans? Try removing "masterpiece, best quality, ....."

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.

Try mixing natural language and booru-styled tags. 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, stronger bias towards females/humans

  • Prompt in the style of photography/non-anime models, "professional photograph of ..... "

    • More realistic images, higher style variety, (probably) less bias

πŸ“™ Model Merging Info - v3.1

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 ones in 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 which inspired me to try mixing my own models.

Thank you to hako-mikan for creating the SuperMerger extension allowing a quicker workflow for merging.

Thank you to those that have contributed to these MBW Rentrys.

Thank you to gsdf, rqdwdw for Counterfeit.

Thank you to RunDiffusion for RunDiffusion FX Photorealistic.

Thank you to andite for Pastel-Mix.

Thank you to DynamicWang for AWPainting.

Thank you to Ayoni for AyoniMix.

Thank you to nuigurumi for BasilMix.

Thank you to KondooAI for Juggernaut.

Thank you to Xerxemi and Xynon for SD-Silicon.

Thank you to AUTO1111 for creating the web UI.

Thank you to StabilityAI for starting everything with Stable Diffusion.

Thank you to the entire SD community for continuing to openly share and create.