This is a weight adjusted through Lokr, which fixes the clip issue and integrates longclip and detail vit. It can be used directly by loading the checkpoint

Using over 1000 images to adjust the model, we have achieved the goal of outputting images of Asian women that are indistinguishable from reality.

Please note that this model does not contain any nsfw content.

Through the training and adjustment of this model, the training principle of flux has been basically understood, and the next step is to carry out large-scale fine-tuning. But after my calculations, this requires a very high level of computing power (economic strength). So the necessary condition is to wait until I have money hahaha.