Flux [dev]lora trained on 64 images of screencaps from a well known children's television show. Characters were gen filled out to avoid training of andromorphic animals into the dataset.
Take the time to look through the sample photos I have provided to get a gauge of the strength and it's influence. Left most photo is no lora, and then it steps 25% up to 100% on the far right.
100% strength is very aggressive with this lora, I prefer 75% or lower.
There was a new patch that made previous LoRas contain artifacts. Not an issue if you are using an upscale pass.
Decided to retrain on Prodigy and I prefer this one. It REALLY captures the dataset source images at full strength, but can be scaled really well to slowly apply.
Recommend not going above 1.0 strength.
SAMPLE IMAGES SHOW EFFECT FROM STRENGTH 0 -1.0 in 0,25 INTERVALS
1. 轉载模型僅供學習與交流分享，其版權及最終解释權归原作者。
2. 模型原作者如需認領模型，請通過官方渠道联系海藝AI工作人員進行認證。我們致力於保護每一位創作者的權益。 點擊去認領
