Iteration 2

This is a decent-ish model if you are using just the checkpoint itself. Where the model shines is copying over details and style from the lora.

In my testing, a strength of 0.8 to 1 works for most lora. For some cartoon focused characters/style, I have gone upto lora strength of 1.2 and it works fine. If you are using realistic style lora then I highly recommend using the detail tweaker betweem 1.2 to 1.5.

https://civitai.com/models/122359/detail-tweaker-xl

Prompting, I use the following format and it works well for me.

angle/camera, score tags, rating/quality tags,

character-lora, character definition,

clothes definition,

pose definition,

background definition, style lora,

Example:

full body, score_9, score_8_up, score_7_up, anime screencap, anime colors,

<lora:kumiko-oumae-s2-ponyxl-lora-nochekaiser:0.95> kumiko oumae, brown eyes, brown hair, short hair, wavy hair, narrowed eyes, tired expression, indifferent, looking to the side, facing viewer, messy hair,

pajamas, morning, sunshine, mood lighting, (dark:1.1),

For cartoon styled loras, experiment removing the score tags completely to get a more authentic style.

You dont need a lot of negatives. Play around to remove unwanted items from the image.

Adetailer is pretty much a must for face fix.