Training checkpoints necessitate a lot of coffee. If you think that what I do is useful and want to help me keep doing this you can buy me a coffe here: https://ko-fi.com/clumsy_trainer
"This is Clazy! a new model I trained using Shivam's repo of Dreambooth. Use "clazy style" (without quotation marks) in prompt for the model to kick-in. At the moment need more finetuning, seems to have problems with mouths. Put "open mouth" in the negative prompt to mitigate this effect. It should work with animals too, but since it was trained on human subjects you need to add brackets. Like this for example: [clazy style:10] where 10 is the sample you want the model to kick-in, the first 10 samples will be dedicated to generate some noise of the actual animal. Also you can use this to generate better likeness on some subjects if you aren't satisfied with the results. Use a CFG scale between 3 and 7 to obtain a smoother model, raise the steps number accordingly (between 20 and 50 should be enough)."