Main prompt tags:
golzineAP, 1boy, old man, anime coloringStyle control:
for the original anime TV show style add following positive prompt: anime coloring
to avoid the style from the anime show, remove anime coloring from positive prompt
Recommended samplers:
Preview1- Euler Ancestral Scheduler: Simple CFG: 4.0 Steps: 30 Lora strength: 1.0
Preview3 - Euler Ancestral Scheduler: Simple CFG: 4.0-.50 Steps: 35-40 Lora strength: 1.0 - x1152 base res helps with wider shots
(er_sde and Euler samplers also work well)
Prompt tips (green - positive, red - negative):
- anime coloring - most important tag to keep the original style of the anime
- black suit, ascot, collared shirt, black pants, dress shoes - base clothes, canonically he wears a lot of different colored suits/ascots/shirts - pick whatever you prefer
- sweater, robe, kimono, bowtie - other canonical clothes
- 6 fingers, six fingers, 4 fingers, four fingers - I think Anima sometimes has issues with too many/too few fingers, this should help it a bit
The LoRa should work with Anima checkpoints. Tested working well with the following:
AnimaYume v0.1 [Preview1]
AnimaYume v0.4 [Preview3]
Last image in the gallery can be imported into ComfyUI with a basic workflow:
No custom nodes required.
Description
I wanted to train Dino for a while but I never liked the results on Illustrious.
SDXL struggles really hard with depicting cowboy shots/full body shots of anything that isn't a 1girl character. Capturing facial details of older male or anthropomorphic characters is super hard without doing HiresFix or facial/eye detailers.
Anima, however, handles it very well. Its not prefect - very wide shots will still mess up the eyes but even so its a massive step-up from Illustrious.