The base is NO MORE Pony. I'm working over a mini own model, merging it with sub-merges of mine+FLUX, mine+Pony, mine+...

Positive prompts are 3 different, all valid since I was too lazy to re-tag so I used datasets with Pony, Illustrious, Danbooru and SDXL regular prompting.

(masterpiece, best quality, high quality, comic, cartoon, toon style, flat color),

(score_9, score_8_up, masterpiece, best quality, high quality, source_cartoon, Toon, flat color),

(score_9_up, score_9, masterpiece, best quality, high quality, cartoon, toon, flat color, western style),

Negative are only:

anime, <Capitan_error>, low quality, worst quality, jpeg artifacts, fused elements, (no_lineart), blurry, bad anatomy, bad quality, worst detail, censor, lowres, watermark, bad quality, worst quality, (manga), <Capitan_animanga>

• No CLIP skip

• 18 ~ 27 steps

• 4.5 ~ 7.5 CFG

• Sampler: euler, DDIM, or Euler_A

• Scheduler: Simple, Karras, or normal

• Datasets had sizes from 768 to 1408

I would ask everyone to NOT make merges of this merge but this would be hypocrite of me, anyway, at least don't monetize LoRas, merges or pictures made with this. That's all.

I merged as "negative" LoRa some anime parameters so it can focus on Toon and Cartoon, if you delete the <Capitan_animanga> you could call some know characters like Vegeta, Saitama, Naruto, Ichigo, etc.

________________________________________________________________________________
________________________________________________________________________________
This is a model made with tons of screenshots from popular series, 51 in western style, with strong toon / comic presence. I used lots of "free to use" pictures, so quality was improved through LoRas merged into model, along with the best compatible VAE I found.
I didn't tag characters, so it won't know who is "Eric Cartman" but can handle many different styles.
I merged into the model (not a Pony) many compatible toon / cartoon / comic LoRas made with Pony, and discovered that if you use scores, 7 out of 10 times, you get a very strong style with a random improve in quality. ¡Bingo!

You can add "NSFW" to get lewd-er outcomes, but a big part of the training set... like... 70% was safe, so it could struggle to make "futa", "facesit" or such.
Compatible with SDXL, Pony and some NoobAI, Hoshino (Pixai) and Haruka (Pixai) LoRas.
For tips and trigger words, check pictures.

Steps: 12 (kind of 2.5D) to 20 (use only in large images)

CFG: 3.5 to 6 (go 0.3 points up every 1 step... approx.)

Samplers and scheduler: in pictures.

Negative: (feel free to add what you want to remove from picture, but removing any of following words will just decrease quality. Tested)

bad quality, worst quality, worst detail, artifacts, signed, censor, (bad hand, bad feet), long chest, chibi, jpeg artifacts, lowres, watermark, real