You can also run this model on sinkin.ai: https://sinkin.ai/m/gLv9zeq
Want to buy me coffee? (Buy a cup)
Updates: 11/23/23 - Uploaded version 11. Please check the "about version" for details.
RealCartoon3D is my first model uploaded. The reason for the upload was to share the fruit of trying to get a checkpoint to do what I wanted it to do when prompted. Some goals for this checkpoint were as follows:
1. Variety in humans (I.E. African, European, Asian, etc). I did not want it just producing the same look I saw everywhere.
2. Produce realistic flaws in skin (just adds to the realism).
3. Do well with LoRAs.
The mission was/is to attempt to get this checkpoint to a point where it will do well on first attempt or second attempt with prompts (my computer is just a gaming laptop that gets really hot when doing things like this lol).
I have learned a lot in the process and even started other checkpoints (RealCartoon-Anime, Realistic, Pixar) to give a more focused variation. This checkpoint being the basis for all of them. I always keep this one merged in all the updates of the others.
I hope you all enjoy it! Please review and share your images. I very much appreciate the support with the downloads and feedback (THANK YOU ALL). Never thought it would get this much attention.
The starting checkpoints for merging were a couple of top ones of course (The checkpoints do/did not have restrictions on checkpoint mergers. These are in my recommended resources as well). I also baked in the VAE (vae-ft-mse-840000-ema-pruned). I tried ClearVAE, but it would mess up from time to time (which may have been my computer). I did not want this issue to fall to anyone that downloaded this model, so I did not use that VAE. Sadly, I did not have the resources to train from scratch; but found that many people would just train off these checkpoints anyway. As time moved on, I would try to find checkpoints that had a look or cool backdrop (or odd result sometimes) that would help the overall look (again avoiding those that had restrictions with merging, as I do not want to mess up anyone's work or get lost in licensing). One issue that would always seem to show up was hands being messed up. They just did not come out right (as many checkpoints seemed to have a problem with); but as I kept moving forward in merging, the hands seemed to get better. I would then look for LoRas to influence the look and style (Ones used are in the resource section). These LoRas were not to take over the checkpoint but help mold it per say. As I wanted the user to have control of this. Since that is the point of LoRas.
Overall, this checkpoint moved up in versions quickly as it got figured out. Eventually it started to really go where I wanted with Version 3.0, 3.1, and then of course Version 4 (slowing down a bit in the updates as well). I still like the older versions; and they did influence the RealCartoon-Pixar and Anime directions.
Width: 512 (flip width and height if doing landscape style image)
Height: 768 (have done up to 1028, but recently been using 904 for portrait style images)
Sampling Method: "Eular A" became pretty much the default as versions moved along, but "DPM++ SDE Karra" and "DPM++ 2M Karras" are favorites for me as well.
Sampling steps: 30-50 normally (30 being my starting point, but going up to 45 a lot of the time). This depending on prompt as well.
Hires.fix settings: Upscaler (R-ESRGAN 4x+, 4k-UltraSharp most of the time), Hires Steps (10-20), Denoising Str (0.34 - 0.45 normally), Upscale (1.5 or 2 does well)
Clip Skip: 2
Would run ADetailer to help enhance the eyes but many times in the later versions I would not run this as it was not needed.
Some settings I run on the web-Ui to help get the images without crashing:
set COMMANDLINE_ARGS=--opt-sdp-no-mem-attention --lowvram --opt-split-attention --xformers
Why So Many Versions:
Because I wanted to share all the results that I felt reached a desired outcome. Allowed me to have fun, and I saw that many enjoyed them. Which motivated me to keep trying. Again, thank you.
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 here Stable Diffusion
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).
- 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.
If you see any conflicts or corrections to be made, please let me know.