ADD_RC_SM v3.0:

(Add RealisticCartoon SuperMerger v3.0)

SEE REWORKED PROCESS BELOW

The Process:

So, here we are again. You, me, and the merge.

I took a different approach this time and did some thing I'm calling a "3rd Phase Injection". I'm sure there's an "official" name for what I did here.. But that's what I'm gonna call it!

The question I had was: "Can I merge 9 models to make one 'Super Merge' without sacrificing prompt accuracy, quality, composition, and least importantly my sanity?". Turns out the answer is yes I can. Not that it didn't take a considerable amount of trial and error. (and one mistake that set me back almost half a day).

So what did I do? Well first let me credit the authors of the amazing checkpoints included in this merge. I, quite literally, couldn't have driven myself to the brink of madness without their hard work. (Even though one of them released an new version in the middle of my merge... @LordTempus lookin' at you!)

I'd also like to take a moment to thank @RestlessDiffusion for their amazing merge guide: [Model Merging Management] - How to merge Stable Diffusion Models to fit your style

That article is highly detailed and gives you a solid foundation and workflow to start your merging journey. Kudos to the Author!

This is a merge of 9 models in 4 Phases:

If you'd prefer to see a technical visualization of the merge scroll to the end.

Phase 1:

• AbsoluteReality

• PerfectPhoton

• AnalogMadness

• RealisticDigitalGenius

• DreamShaper

• dpsRealisticMix

Phase 2:

• LiangyiusRealistic

• UltraFantasy

Phase 3:

• RealCartoon-Realistic

Phase 4:

• Adjusted Noise, Contrast, Brightness, and Color Composition

All great Checkpoints and Creators that you should definitely check out if you haven't already!

So now the "nitty-gritty" as it were:

As shown above, the merge was done in 4 phases. With Phase 4 being an Adjustment Phase.

First phase was to create the "Base", or as I call it in the merge doc, Part1.

The second phase was to create the second part of the 'Base' or Part2.

Why did I do it this way? Composition; I paired each model as 'like pairs'. Then matched those pairs with 'like groups'.

The Phase 1 Group all share compositional similarities that complimented the others. The two outliers were LiangyiusRealistic and UltraFantasy.

Both those Checkpoints have an awesome look but they don't merge well with the Phase 1 Group; at least not directly.

Phase 3 consisted of Part1, Part2, and RealCartoon-Realistic being merged together. I had originally merged Part1 and Part2 then merged-in RC-R, but the result was less than pleasing for my intention. So what to do.

I tried merging RC-R into either Part1 and Part2 independently, but then the composition got lost when I merged those two parts together.

So I then tried merging RC-R into both Part1 and Part2 and then merging those two parts. But again, it was not to my liking, and I was starting to loose composition. (and my sanity at this point.)

The last thing I could do was merge Part1, part2, and RC-R directly. But I couldn't just merge them, I had to mix them in such a way that the best parts of each shine in the final merge.

After a few hours of 'experimenting' I finally landed on what we have here.

ADD_RC_SM v3.0

Give It a Go! All Opinions Welcome!

Full Res Img HERE

Recommended Settings:

• Width: 512 (768)

• Height: 768 (1536)

• Sampling Method: DPM++ 3M SDE Exponential is my "go to". 'Euler a' works pretty well too.

• Sampling steps: 30 - 90 (40-60 is the sweet spot)

• Clip Skip: 2

• Upscaling: Ultimate SD Upscale - 4xUltraSharp - 1.5 to 2 perpetual

I recommend using ADetailer to enhance the World, Eyes, Face, Hands, Feet, and Clothing ONLY. The Model is semi-high contrast and already includes A LOT of detail. If you try to enhance the body you will end up getting a bad result.

License & Use

This model is open access and available to all, with a CreativeML OpenRAIL-M license further specifying rights and usage.

• 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. 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 modified CreativeML OpenRAIL-M to all your users (please read the license entirely and carefully).

  Please read the full license.

Use Restrictions:

You agree not to use the Model or Derivatives of the Model:

- In any way that violates any applicable national, federal, state, local or international law or regulation

- For the purpose of exploiting, harming or attempting to exploit or harm minors in any way

- To generate or disseminate verifiably false information and/or content with the purpose of harming others

- To generate or disseminate personal identifiable information that can be used to harm an individual

- To defame, disparage or otherwise harass others

- For fully automated decision making that adversely impacts an individual’s legal rights or otherwise creates or modifies a binding, enforceable obligation

- For any use intended to or which has the effect of discriminating against or harming individuals or groups based on online or offline social behavior or known or predicted personal or personality characteristics

- To exploit any of the vulnerabilities of a specific group of persons based on their age, social, physical or mental characteristics, in order to materially distort the behavior of a person pertaining to that group in a manner that causes or is likely to cause that person or another person physical or psychological harm

- For any use intended to or which has the effect of discriminating against individuals or groups based on legally protected characteristics or categories

- To provide medical advice and medical results interpretation

- To generate or disseminate information for the purpose to be used for administration of justice, law enforcement, immigration or asylum processes, such as predicting an individual will commit fraud/crime commitment (e.g. by text profiling, drawing causal relationships between assertions made in documents, indiscriminate and arbitrarily-targeted use).

Terms of use:

- You are solely responsible for any legal liability resulting from unethical use of this model(s)

- If you use any of these models for merging, please state what steps you took to do so and clearly indicate where modifications have been made.