Hi, I wanna talk a little bit about my models. Especially my SDXL-based models.
As you know, I've been very active in developing and maintaining my SDXL-based models recently. To be honest, I'm kinda sad that my SDXL-based models can hardly get 1k downloads. I mean, some models that used my model on their merge even get more downloads than mine. It's okay, maybe my SDXL work isn't the best. But hey, I tried my best with all of my limitations.
On the other side, I have been very tired of handling my SD 1.5-based models. There is just nothing to improve, it's just stuck like that. No matter how much I've done with it, there are not many improvements there. That's why I am considering stopping developing my SD 1.5-based models.
If I could be honest with you, if I post and train models only for fame, I would just stuck with SD 1.5-based models forever. It's cheaper to develop because I can use my own hardware. But no, I choose to burn my monthly pocket money to train SDXL on Runpod.
I have two requests, one for my model users, and one for model trainers who use my models to train.
I know that not everyone has the hardware powerful enough to run SDXL, that's why I've asked the Civitai team to turn on on-site generations on both my SDXL-based models. Please use my models with it. By only using it, you will help me, a lot.
If you're merging or training with my models, please cite my models correctly. It will make things easier to see who is using my models for their development. I might use your fine-tuned version of my models for my development too, as sefuh (senior) Linaqruf said:
Open-source models are the power of everyone. We can make better things by democratizing generative AI models.
That's all I can say. Thank you for reading.