uwustyle - Chroma Edition

Details

Download Files

About this version

Combined the training datasets of v1 and v2 and then reduced down to the 86 best looking images, as I am aiming for a look somewhere between the two. I also added Illustrious-style aesthetic quality tags to approximate half of the training data (the better half).

Changed training parameters:

Batch size: 1 -> 2
UNet LR: 0.25 -> 0.5 (prodigy)
Dim/Alpha: 48/48 -> 64/64
Steps: 3600 -> 8600

Model description

Trained on edited cute furry & kemono illustrations + cartoon illustrations of animals + upscaled & cropped anime frames, captioned with both E621 tags and natural language of varying verbosity.

Prompt with primarily E621 tags and natural language wherever you want to be more specific, in any order desired.

I personally generate at SD1.5 resolutions (Chroma 1.0-HD is trained at both SD1.5 and SDXL sizes), upscale to the actual resolution I want, and then resample again for 14-19 steps. I also use the flan-t5-xxl CLIP model rather than the usual one.

Images made by this model

Sort by

No Images Found.

Model Type	LORA
Base Model	Chroma
Published	10/22/2025
Trained Words	anthro animal focus cute high quality, masterpiece