This model is intended to build upon Lianima to add bold lines and better image composition, in order to personally use later it for pixel/game art.

In order to do so, here is how it was made:

• As a base, i leveraged Diving-Anima V9 by @DivingSuit instead of base ANIMA.

• I converted it to ANIMA 2.9B using my tool.

• I merged the source LoRA (unpublished) of Lianima.

• I adapted the great LineWeight LoRA by @ABridgetooFar (new account of BridgeToons, give them some love) => scaled to 1.1, converted to 2.9B, "compressed" using my tool for Rank.

• I generated ~1000 pictures and selected ~600 i prefered to introduce my own bias 💜

• Trained for 4 epochs (~2500 steps) a LoRA based on those pictures.

• Merged it into the resulting checkpoint, together with Turbo V4.

For usage:

• CFG ~1.2, Steps 16, Hires Steps 16, Hires denoise 0.3, Hires x1.5, 4xNomos8kSC

• ER_SDE + Beta

• Flow Shift 3

• Positive: masterpiece, best quality, highres, <your prompt>

• Negative: worst quality, low quality, blurry, anatomical nonsense, text, watermark, signature, child, loli

Special feature: Contains my OC Liadora baked in:

• Liadora, 1girl, short red hair, green eyes, black horns, liadoraoutfit, black pleated latex dress, purple elbow gloves, thighhigh boots

Merge is authorized, but a link to this model card is required i you do so.

Style comparaison with Lianima here: https://civarchive.com/images/147193556