No trigger word is needed for this LoRA. Though using something like This is a absurdres, high quality, aesthetic masterpiece cinematic anime artwork. or something around those lines helps Chroma produce better images.
I invite you to add other LoRA to make the image quality better. The showcase images were generated with a very basic workflow (comfy workflow is included in the images). You can easily get better result if you try a bit harder, this is only meant to showcase the effect of the LoRA alone.
I recommend setting the LoRA strength at 0.9 for the best results.
V2 was trained in Onetrainer using Prodigy Plus Schedule Free at batch size 4 for 5500 steps at 512px (I might do a v2.1 at 1024px if my GPU has nothing to do for ~16h).
I added a few images and did a better job cropping than the first time.
Captioning was done Qwen3-VL-8B-NSFW-Caption-V4.5 which is better than joycaption at everything but hardcore content.
Description
Lora trained on 60 PossumMachine images at 1MP, using kohya-ss/sd-scripts with Adamw -1e04 LR constant for ~5k steps
There is a lot of room for improvement by changing the LR to something smaller at the end, adding a reg dataset, masking image with white background, ...
The captioning is also mediocre at best (JoyCaption makes a LOT of mistakes).