This is an experiment to test whether the LoRA (trained with a series of illustrations emphasizing shadow rendering, color composition, and poetic philosophy) can reproduce the aforementioned effects in SDXL.
(qwen image version is a test in test.)
Model name: 未竟歌诗 / Incomplete Poetry
Use this LoRA to bring fresh, elegant, and tranquil aesthetic tones to your images, or an intense, unique, and profound atmosphere!
It would bring a very volumetric and natural style to clouds, which means it would decrease the ai-generated style.
while using the qwen image version, add these into prompt:
((2D illustration, pixiv illustration, trending on pixiv)),
it'll help you get more painterly results.
the weight of qwen version should be 1.0-1.4
steps need to be over 25+, though it's expensive
In illustrious/noob ai version
Recommended parameters: Enable in A1111
Recommended weight: 1.0 (less effect with lower weight. some base models can pull the weight to 1.5–2.0 without overfitting)
clipskip=2