Style > People's Work
Details
Download Files
Model description
No trigger word needed for this lora. It should impove the quality of images generated for anime and digital illustration overall.
You should pair it with another lora to steer Chroma towards the style you want your image to have, this one should only aims to increase quality.
The very basic comfy workflow is available in the images. You can get better result both with and without the lora with a more complexe one (res_2m + bong_tangent would be a good start)
Lora trained at rank 16 on 1000 hand picked high quality images posted on civitai at 1MP, using OneTrainer, with Adamw, -1e04 LR constant, batch size 4, for 11 epochs.
There is a room for improvement by changing the LR to something smaller at the end, adding a reg dataset, ...
The captioning is also mediocre at best (JoyCaption makes a LOT of mistakes).
I might do a V2 once I have a better pipeline going.