zMix - TooniVerse XL

zMix - TooniVerse XL

TooniVerse, now in PonyXL!

It's been a while coming, and probably I am way to late for the party- but here it is. TooniVerse XL. The model is based on Pony XL, meaning that it prefers the "score" prompt:

score_9, score_8_up, score_7_up, score_6_up, [etc]

This merge is made using the following (Beta 1):

• AutismMix Pony

• A custom mix of Animerge v3.1 and SeaArt Furry v1

• AnimeConfettiTuneXL

The following LORA:

• add-detail-xl

• BetterBulge

• Hotcha

• BoxerHole Style

• Asura Style

• Legends of Runeterra

Beta 2 consists out of:

• AutismMix Confetti

• Animagine XL3.1

• SeaArt Furry v1

  (more to come, mind = foggy)

The following LORA have been merged:

• ExpressiveH

• BetterBulge

• Add-detail-XL

• Legends of Runeterra

• Vixon's Pony Style - Emotive

• Vixon's Pony Style - Facialhair

• Coral Island Style (Different epoch curtesy of Barracuda)

Recommended settings:

I've mainly tested the merge with the settings below- this is largely because I've done most of the process in ComfyUI. By all means experiment!

• Sampler: Euler A or DPM++ 2M Karras

• Steps: 40 (beta 1), 30/40 (beta 2)

• Hires Fix: 4x_NMKD-Superscale, 2x, 0.45 denoise

• Size: 1156x896/896x1156, 1216x832/832x1216, 1200x1200

• CFG: 7 / 8 (beta 1), 6/7 (beta 2)

• Clipskip: 2 (though, not needed I have it set to 2 anyway)

• VAE: baked in, but expierment

• Adetailer: face_yolov8n_v2 and mediapike_face_mesh_eyes_only

Known quirks:

• The model at time really is keen on placing watermarks onto the pictures- I am assuming this is because of the data that Lora has been trained with; I will try to fix it in future versions. (Beta 1, Beta 2 does not have this issue)

Special Thanks:

• TheHiddenOne - For providing me with earlier Epochs of his model so I could merge his style into the Model.

• Invariable - Still being an amazing tester and friend

• Scratchproof - Testing the model and keeping me (in)sane with other mischief