Magical Girls Football Club Mix版本MGFC_Mix_V-1.1.2 (ID: 9115)

Welcome to the Magical Girls Football Club!

This is a niche mix/merge model I've been working on recently for the purpose of generating anime style magical girls playing American football/wearing football uniforms. This is the area I started specializing in when stable diffusion started to take off due to being a fan of both anime and football after doing random generations with topics/themes that didn't fit with one another. Another reason for choosing this topic to dive into is as there is little to no preexisting content which easily allows me to see how the tech advances through my generations and seeing how specialized I can get.

This model is what I've found to be the best at generating images in my style/niche so far after experimenting with many, many different models, but it is definitely a work in progress. As I continue to experiment with this model, I will add updated versions and documentation of any changes or updates.

Things that are currently giving me problems with generations are:

1. Eye black - This is my white whale, as of now nothing has produced eye black consistently or hardly ever.

2. Uniform numbering - As with any text/lettering in SD generation this is just a crapshoot of what you're going to get, but it seems to be getting a bit more consistent.

3. Consistent football shape - Getting better!

4. General AI generation issues - Hands not amazing, faces can be rough, etc

If anyone has any feedback or suggestions, feel free to reach out and let me know, especially if anyone knows any method for adding eye black generation to my mix! I am still very new to this, so it goes without saying I don't know what the hell I'm doing, so be patient with me if the model does not perform as expected.

... Wait what do you mean how does it preform for regular anime generations, why would you wanna do that?

The model will work just like any other standard anime SD model and should provide pretty decent results as this is just a mix of without any specific training done to skew it.

描述:

Initial release model version!

Recommended Settings are:

• Sampler: DPM++ 2S a Karras

• Steps: 25 - 50

• CFG: 7.5 - 13

• Clip Skip - 2

• R-ESRGAN 4x+ Anime6B

  • Hires Steps: 10 - 15

  • Denoising Strength: 0.55 - 0.69

  • Upscale: 2

Mix of DreamShaper 3.32, Sailor Moon Mix, and Pastel-Mix with the Pastel-Waifu VAE baked in. Don't have the exact weight formula available as of now.

训练词语: